Anthropic, maker of the chatbot Claude, declined to comment on the meeting but CEO Dario Amodei has made clear his ethical concerns about unchecked government use of AI, including the dangers of fully autonomous armed drones and of AI-assisted mass surveillance that could track dissent.

The meeting between Hegseth and Amodei was confirmed by a defense official who was not authorized to comment publicly and spoke on condition of anonymity.

It underscores the debate over AI’s role in national security and concerns about how the technology could be used in high-stakes situations involving lethal force, sensitive information or government surveillance. It also comes as Hegseth has vowed to root out what he calls a “woke culture” in the armed forces.

“A powerful AI looking across billions of conversations from millions of people could gauge public sentiment, detect pockets of disloyalty forming, and stamp them out before they grow,” Amodei wrote in an essay last month.

Anthropic is the only AI company approved for classified military networks

The Pentagon announced last summer that it was awarding defense contracts to four AI companies — Anthropic, Google, OpenAI and Elon Musk’s xAI. Each contract is worth up to $200 million.

Anthropic was the first AI company to get approved for classified military networks, where it works with partners like Palantir. The other three companies, for now, are only operating in unclassified environments.

By early this year, Hegseth was highlighting only two of them: xAI and Google.

The defense secretary said in a January speech at Musk’s space flight company, SpaceX, in South Texas that he was shrugging off any AI models “that won’t allow you to fight wars.”

Hegseth said his vision for military AI systems means that they operate “without ideological constraints that limit lawful military applications,” before adding that the Pentagon’s “AI will not be woke.”

In January, Hegseth said Musk’s artificial intelligence chatbot Grok would join the Pentagon network, called GenAI.mil. The announcement came days after Grok — which is embedded into X, the social media network owned by Musk — drew global scrutiny for generating highly sexualized deepfake images of people without their consent.

OpenAI announced in early February that it, too, would join the military’s secure AI platform, enabling service members to use a custom version of ChatGPT for unclassified tasks.

Anthropic calls itself more safety-minded

Anthropic has long pitched itself as the more responsible and safety-minded of the leading AI companies, ever since its founders quit OpenAI to form the startup in 2021.

The uncertainty with the Pentagon is putting those intentions to the test, according to Owen Daniels, associate director of analysis and fellow at Georgetown University’s Center for Security and Emerging Technology.

“Anthropic’s peers, including Meta, Google and xAI, have been willing to comply with the department’s policy on using models for all lawful applications,” Owens said. “So the company’s bargaining power here is limited, and it risks losing influence in the department’s push to adopt AI.”

In the AI craze that followed the release of ChatGPT, Anthropic closely aligned with President Joe Biden’s administration in volunteering to subject its AI systems to third-party scrutiny to guard against national security risks.

Amodei, the CEO, has warned of AI’s potentially catastrophic dangers while rejecting the label that he’s an AI “doomer.” He argued in the January essay that “we are considerably closer to real danger in 2026 than we were in 2023″ but that those risks should be managed in a “realistic, pragmatic manner.”

Anthropic has been at odds with the Trump administration

This would not be the first time Anthropic’s advocacy for stricter AI safeguards has put it at odds with the Trump administration. Anthropic needled chipmaker Nvidia publicly, criticizing Trump’s proposals to loosen export controls to enable some AI computer chips to be sold in China. The AI company, however, remains a close partner with Nvidia.

The Trump administration and Anthropic also have been on opposite sides of a lobbying push to regulate AI in U.S. states.

Trump’s top AI adviser, David Sacks, accused Anthropic in October of “running a sophisticated regulatory capture strategy based on fear-mongering.”

Sacks made the remarks on X in response to an Anthropic co-founder, Jack Clark, writing about his attempt to balance technological optimism with “appropriate fear” about the steady march toward more capable AI systems.

Anthropic hired a number of ex-Biden officials soon after Trump’s return to the White House, but it’s also tried to signal a bipartisan approach. The company recently added Chris Liddell, a former White House official from Trump’s first term, to its board of directors.

The Pentagon-Anthropic debate is reminiscent of an uproar several years ago when some tech workers objected to their companies’ participation in Project Maven, a Pentagon drone surveillance program. While some workers quit over the project and Google itself dropped out, the Pentagon’s reliance on drone surveillance has only increased.

Similarly, “the use of AI in military contexts is already a reality and it is not going away,” Owens said.

“Some contexts are lower stakes, including for back-office work, but battlefield deployments of AI entail different, higher-stakes risks,” he said, referring to the use of lethal force or weapons like nuclear arms. “Military users are aware of these risks and have been thinking about mitigation for almost a decade.”

“When government agencies use AI tools, we will now require them to verify that those tools do not endanger the rights and safety of the American people,” Vice President Kamala Harris told reporters ahead of the announcement.

Each agency by December must have a set of concrete safeguards that guide everything from facial recognition screenings at airports to AI tools that help control the electric grid or determine mortgages and home insurance.

The new policy directive being issued to agency heads Thursday by the White House’s Office of Management and Budget is part of the more sweeping AI executive order signed by President Joe Biden in October.

While Biden’s broader order also attempts to safeguard the more advanced commercial AI systems made by leading technology companies, such as those powering generative AI chatbots, Thursday’s directive will also affect AI tools that government agencies have been using for years to help with decisions about immigration, housing, child welfare and a range of other services.

As an example, Harris said, “If the Veterans Administration wants to use AI in VA hospitals to help doctors diagnose patients, they would first have to demonstrate that AI does not produce racially biased diagnoses.”

Agencies that can’t apply the safeguards “must cease using the AI system, unless agency leadership justifies why doing so would increase risks to safety or rights overall or would create an unacceptable impediment to critical agency operations,” according to a White House announcement.

The new policy also calls for two other “binding requirements,” Harris said. One is that federal agencies must hire a chief AI officer with the “experience, expertise and authority” to oversee all of the AI technologies used by that agency, she said. The other is that each year, agencies must make public an inventory of their AI systems that includes an assessment of the risks they might pose.

Some rules exempt intelligence agencies and the Department of Defense, which is having a separate debate about the use of autonomous weapons.

Shalanda Young, the director of the Office of Management and Budget, said the new requirements are also meant to strengthen positive uses of AI by the U.S. government.

“When used and overseen responsibly, AI can help agencies to reduce wait times for critical government services, improve accuracy and expand access to essential public services,” Young said.

The new oversight was applauded Thursday by civil rights groups, some of which have spent years pushing federal and local law enforcement agencies to curb the use of face recognition technology tied to wrongful arrests of Black men.

A September report by the U.S. Government Accountability Office reviewing seven federal law enforcement agencies, including the FBI, found that they cumulatively conducted more than 60,000 searches using face-scanning technology without first requiring sufficient staff training on how it works and how to interpret results.

The order reflects the government’s effort to shape how AI evolves in a way that can maximize its possibilities and contain its perils. AI has been a source of deep personal interest for Biden, with its potential to affect the economy and national security.

White House chief of staff Jeff Zients recalled Biden giving his staff a directive to move with urgency on the issue, having considered the technology a top priority.

“We can’t move at a normal government pace,” Zients said the Democratic president told him. “We have to move as fast, if not faster than the technology itself.”

In Biden’s view, the government was late to address the risks of social media and now U.S. youth are grappling with related mental health issues. AI has the positive ability to accelerate cancer research, model the impacts of climate change, boost economic output and improve government services among other benefits. But it could also warp basic notions of truth with false images, deepen racial and social inequalities and provide a tool to scammers and criminals.

The order builds on voluntary commitments already made by technology companies. It’s part of a broader strategy that administration officials say also includes congressional legislation and international diplomacy, a sign of the disruptions already caused by the introduction of new AI tools such as ChatGPT that can generate new text, images and sounds.

Using the Defense Production Act, the order will require leading AI developers to share safety test results and other information with the government. The National Institute of Standards and Technology is to create standards to ensure AI tools are safe and secure before public release.

The Commerce Department is to issue guidance to label and watermark AI-generated content to help differentiate between authentic interactions and those generated by software. The order also touches on matters of privacy, civil rights, consumer protections, scientific research and worker rights.

An administration official who previewed the order on a Sunday call with reporters said the to-do lists within the order will be implemented and fulfilled over the range of 90 days to 365 days, with the safety and security items facing the earliest deadlines. The official briefed reporters on condition of anonymity, as required by the White House.

Last Thursday, Biden gathered his aides in the Oval Office to review and finalize the executive order, a 30-minute meeting that stretched to 70 minutes, despite other pressing matters including the mass shooting in Maine, the Israel-Hamas war and the selection of a new House speaker.

Biden was profoundly curious about the technology in the months of meetings that led up to drafting the order. His science advisory council focused on AI at two meetings and his Cabinet discussed it at two meetings. The president also pressed tech executives and civil society advocates about the technology’s capabilities at multiple gatherings.

“He was as impressed and alarmed as anyone,” deputy White House chief of staff Bruce Reed said in an interview. “He saw fake AI images of himself, of his dog. He saw how it can make bad poetry. And he’s seen and heard the incredible and terrifying technology of voice cloning, which can take three seconds of your voice and turn it into an entire fake conversation.”

The possibility of false images and sounds led the president to prioritize the labeling and watermarking of anything produced by AI. Biden also wanted to thwart the risk of older Americans getting a phone call from someone who sounded like a loved one, only to be scammed by an AI tool.

Meetings could go beyond schedule, with Biden telling civil society advocates in a ballroom of San Francisco’s Fairmont Hotel in June: “This is important. Take as long as you need.”

The president also talked with scientists and saw the upside that AI created if harnessed for good. He listened to a Nobel Prize-winning physicist talk about how AI could explain the origins of the universe. Another scientist showed how AI could model extreme weather like 100-year floods, as the past data used to assess these events has lost its accuracy because of climate change.

The issue of AI was seemingly inescapable for Biden. At Camp David one weekend, he relaxed by watching the Tom Cruise film “Mission: Impossible — Dead Reckoning Part One.” The film’s villain is a sentient and rogue AI known as “the Entity” that sinks a submarine and kills its crew in the movie’s opening minutes.

“If he hadn’t already been concerned about what could go wrong with AI before that movie, he saw plenty more to worry about,” said Reed, who watched the film with the president.

With Congress still in the early stages of debating AI safeguards, Biden’s order stakes out a U.S. perspective as countries around the world race to establish their own guidelines. After more than two years of deliberation, the European Union is putting the final touches on a comprehensive set of regulations that targets the riskiest applications for the technology. China, a key AI rival to the U.S., has also set some rules.

U.K. Prime Minister Rishi Sunak also hopes to carve out a prominent role for Britain as an AI safety hub at a summit this week that Vice President Kamala Harris plans to attend.

The U.S., particularly its West Coast, is home to many of the leading developers of cutting-edge AI technology, including tech giants Google, Meta and Microsoft and AI-focused startups such as OpenAI, maker of ChatGPT. The White House took advantage of that industry weight earlier this year when it secured commitments from those companies to implement safety mechanisms as they build new AI models.

But the White House also faced significant pressure from Democratic allies, including labor and civil rights groups, to make sure its policies reflected their concerns about AI’s real-world harms.

The American Civil Liberties Union is among the groups that met with the White House to try to ensure “we’re holding the tech industry and tech billionaires accountable” so that algorithmic tools “work for all of us and not just a few,” said ReNika Moore, director of the ACLU’s racial justice program.

Suresh Venkatasubramanian, a former Biden administration official who helped craft principles for approaching AI, said one of the biggest challenges within the federal government has been what to do about law enforcement’s use of AI tools, including at U.S. borders.

“These are all places where we know that the use of automation is very problematic, with facial recognition, drone technology,” Venkatasubramanian said. Facial recognition technology has been shown to perform unevenly across racial groups, and has been tied to mistaken arrests.

Outside watchdog groups said misinformation about the war abounds on the platform, whose workforce — including its content moderation team — was gutted by billionaire Elon Musk after he bought it last year.

Fake and manipulated imagery circulating on X include “repurposed old images of unrelated armed conflicts or military footage that actually originated from video games,” said a Tuesday letter to Musk from European Commissioner Thierry Breton. “This appears to be manifestly false or misleading information.”

Breton, the EU’s digital rights chief, also warned Musk that authorities have been flagging “potentially illegal content” that could violate EU laws and “you must be timely, diligent and objective” in removing it when warranted.

San Francisco-based X didn’t immediately respond to a request for comment about Breton’s letter.

A post late Monday from X’s safety team claimed it is treating the crisis with utmost effort: “In the past couple of days, we’ve seen an increase in daily active users on @X in the conflict area, plus there have been more than 50 million posts globally focusing on the weekend’s terrorist attack on Israel by Hamas. As the events continue to unfold rapidly, a cross-company leadership group has assessed this moment as a crisis requiring the highest level of response.”

That includes continuing a policy frequently championed by Musk of letting users help rate what might be misinformation, which causes those posts to include a note of context but not disappear from the platform.

The struggle to identify reliable sources for news about the war was exacerbated over the weekend by Musk, who on Sunday posted the names of two accounts he said were “good” for “following the war in real-time.” Analyst Emerson Brooking of the Atlantic Council called one of those accounts “absolutely poisonous.” Journalists and X users also pointed out that both accounts had previously shared a fake AI-generated image of an explosion at the Pentagon, and that one of them had posted numerous antisemitic comments in recent months. Musk later deleted his post.

Brooking posted on X that Musk had enabled fake war reporting by abandoning the blue check verification system for trusted accounts and allowing anyone to buy a blue check.

Brooking said Tuesday that it is “significantly harder to determine ground truth in this conflict as compared to Russia’s invasion of Ukraine” last year and “Elon Musk bears personal responsibility for this.”

He said Musk’s changes to the X platform have made it impossible to quickly assess the credibility of accounts while his “introduction of view monetization has created perverse incentives for war-focused accounts to post as many times as possible, even unverified rumors, and to make the most salacious claims possible.”

“War is always a cauldron of tragedy and disinformation; Musk has made it worse,” he added. Further, Brooking said via email “Musk has repeatedly and purposefully denigrated the idea of an objective media, and he made platform design decisions that undermine such reporting. We now see the result.”

Part of Musk’s drastic changes over the past year included removing many of the people responsible for moderating toxic content and harmful misinformation.

One former member of Twitter’s public policy team said the company is having a harder time taking action on posts that violate its policies because there aren’t enough people to do that work.

“The layoffs are undermining the capacity of Twitter’s trust and safety team, and associated teams like public policy, to provide needed support during a critical time of crisis,” said Theodora Skeadas, one of thousands of employees who lost their jobs in the months after Musk bought the company.

X says it changed one policy over the weekend to enable people to more easily choose whether or not to see sensitive media without the company actually taking down those posts.

“X believes that, while difficult, it’s in the public’s interest to understand what’s happening in real time,” its statement said.

The company said it is also removing newly created Hamas-affiliated accounts and working with other tech companies to try to prevent “terrorist content” from being distributed online. It said it is “also continuing to proactively monitor for antisemitic speech as part of all our efforts. Plus we’ve taken action to remove several hundred accounts attempting to manipulate trending topics.”

Linda Yaccarino, whom Elon Musk named in May as the top executive at X, withdrew from an upcoming three-day tech conference where she was scheduled to speak, citing the need to focus on how the platform was handling the war.

“With the global crisis unfolding, Linda and her team must remain fully focused on X platform safety,” X told the organizers of the WSJ Tech Live conference being held next week in Laguna Beach, California.

The closed-door forum on Capitol Hill will include almost two dozen tech leaders and advocates, and some of the industry’s biggest names: Meta’s Mark Zuckerberg and X and Tesla’s Elon Musk as well as former Microsoft CEO Bill Gates. All 100 senators are invited, but the public is not.

Schumer, D-N.Y., who’s leading the forum with Republican Sen. Mike Rounds of South Dakota, won’t necessarily take the tech executives’ advice as he works with Republicans and fellow Democrats to try and ensure some oversight of the burgeoning sector. But he’s hoping that they will give senators some realistic direction as he tries to do what Congress hasn’t done for many years — pass meaningful regulation of the tech industry.

“It’s going to be a fascinating group because they have different points of view,” Schumer said in an interview with The Associated Press ahead of the forum. “Hopefully we can weave it into a little bit of some broad consensus.”

Rounds, who spoke to AP with Schumer on Tuesday, said Congress needs to get ahead of fast-moving AI by making sure it continues to develop “on the positive side” while also taking care of potential issues surrounding data transparency and privacy.

“AI is not going away, and it can do some really good things or it can be a real challenge,” Rounds said.

Schumer says regulation of artificial intelligence will be “one of the most difficult issues we can ever take on,” and ticks off the reasons why: It’s technically complicated, it keeps changing and it “has such a wide, broad effect across the whole world,” he said.

Congress has a lackluster track record when it comes to regulating technology. Lawmakers have lots of proposals — many of them bipartisan — but have mostly failed to agree on major legislation to regulate the industry as powerful tech companies have resisted.

Many lawmakers point to the failure to pass any legislation surrounding social media — bills have stalled in both chambers that would better protect children, regulate activity around elections and mandate stricter privacy standards, among other measures.

“We don’t want to do what we did with social media, which is let the techies figure it out, and we’ll fix it later,” says Senate Intelligence Committee Chairman Mark Warner, D-Va., on the AI push.

Schumer’s bipartisan working group — comprised of Rounds, Democratic Sen. Martin Heinrich of New Mexico and Republican Sen. Todd Young of Indiana — is hoping that the rapid growth of artificial intelligence will create more urgency. Sparked by the release of ChatGPT less than a year ago, businesses across many sectors have been clamoring to apply new generative AI tools that can compose human-like passages of text, program computer code and create novel images, audio and video. The hype over such tools has accelerated worries over its potential societal harms and prompted calls for more transparency in how the data behind the new products is collected and used.

“You have to have some government involvement for guardrails,” Schumer said. “If there are no guardrails, who knows what could happen.”

Schumer says Wednesday’s forum will focus on big ideas like whether the government should be involved at all, and what questions Congress should be asking. Each participant will have three minutes to speak on a topic of their choosing, and Schumer and Rounds will moderate open discussions among the group in the morning and afternoon.

Some of Schumer’s most influential guests, including Musk and Sam Altman, CEO of ChatGPT-maker OpenAI, have signaled more dire concerns evoking popular science fiction about the possibility of humanity losing control to advanced AI systems if the right safeguards are not in place.

But for many lawmakers and the people they represent, AI’s effects on employment and navigating a flood of AI-generated misinformation are more immediate effects.

A recent report from the market research group Forrester projected that generative AI technology could replace 2.4 million jobs in the U.S. by 2030, many of them white-collar roles not affected by previous waves of automation. This year alone the number of lost jobs could total 90,000, the report said, though far more jobs will be reshaped than eliminated.

AI experts have also warned of the growing potential of AI-generated online disinformation to influence elections, including the upcoming 2024 presidential race.

On the more positive side, Rounds says he would like to see the empowerment of new medical technologies that could save lives and allow medical professionals to access more data. That topic is “very personal to me,” Rounds says, after his wife died of cancer two years ago.

Many members of Congress agree that legislation will probably be needed in response to the quick escalation of artificial intelligence tools in government, business and daily life. But there is little consensus on what that should be, or what might be needed. There is also some division — some members worry more about overregulation, and others worry more about the potential risks of an unchecked industry.

“I am involved in this process in large measure to ensure that we act, but we don’t act more boldly or over-broadly than the circumstances require,” says Sen. Young, one of the members of Schumer’s working group. “We should be skeptical of government, which is why I think it’s important that you got Republicans at the table.”

Young says that Schumer has reassured him that he will be “hypersensitive to overshooting as we address some of the potential harms of AI.”

Some Republicans have been wary of following the path of the European Union, which signed off in June on the world’s first set of comprehensive rules for artificial intelligence. The EU’s AI Act will govern any product or service that uses an AI system and classify them according to four levels of risk, from minimal to unacceptable.

A group of corporations has called on EU leaders to rethink the rules, arguing that it could make it harder for companies in the 27-nation bloc to compete with rivals overseas in the use of generative AI.

In the United States, most major tech companies have expressed support for AI regulations, though they don’t necessarily agree on what that means.

“We’ve always said that we think that AI should get regulated,” said Dana Rao, general counsel and chief trust officer for software company Adobe. “We’ve talked to Europe about this for the last four years, helping them think through the AI Act they’re about to pass. There are high-risk use cases for AI that we think the government has a role to play in order to make sure they’re safe for the public and the consumer.”

Adobe, which makes Photoshop and the new AI image-generator Firefly, is proposing its own federal legislation: an “anti-impersonation” bill to protect artists as well as AI developers from the misuse of generative AI tools to produce derivative works without a creator’s consent.

Senators say they will figure out a way to regulate the industry, despite the odds.

“Make no mistake. There will be regulation. The only question is how soon, and what,” said Sen. Richard Blumenthal, D-Conn., at a Tuesday hearing on legislation he wrote with Republican Sen. Josh Hawley of Missouri.

Blumenthal’s framework calls for a new “licensing regime” that would require tech companies to seek licenses for high-risk AI systems. It would also create an independent oversight body led by experts and hold companies liable when their products breach privacy or civil rights or endanger the public.

“Risk-based rules, managing the risks, is what we need to do here,” Blumenthal said.

“As this technology advances, we understand that people are anxious about how it could change the way we live. We are too,” OpenAI CEO Sam Altman testified at a Senate hearing Tuesday.

Altman proposed the formation of a U.S. or global agency that would license the most powerful AI systems and have the authority to “take that license away and ensure compliance with safety standards.”

His San Francisco-based startup rocketed to public attention after it released ChatGPT late last year. ChatGPT is a free chatbot tool that answers questions with convincingly human-like responses.

What started out as a panic among educators about ChatGPT’s use to cheat on homework assignments has expanded to broader concerns about the ability of the latest crop of “generative AI” tools to mislead people, spread falsehoods, violate copyright protections and upend some jobs.

And while there’s no immediate sign that Congress will craft sweeping new AI rules, as European lawmakers are doing, the societal concerns brought Altman and other tech CEOs to the White House earlier this month and have led U.S. agencies to promise to crack down on harmful AI products that break existing civil rights and consumer protection laws.

Sen. Richard Blumenthal, the Connecticut Democrat who chairs the Senate Judiciary Committee’s subcommittee on privacy, technology and the law, opened the hearing with a recorded speech that sounded like the senator, but was actually a voice clone trained on Blumenthal’s floor speeches and reciting a speech written by ChatGPT after he asked the chatbot to compose his opening remarks.

The result was impressive, said Blumenthal, but he added, “What if I had asked it, and what if it had provided, an endorsement of Ukraine surrendering or (Russian President) Vladimir Putin’s leadership?”

Blumenthal said AI companies ought to be required to test their systems and disclose known risks before releasing them, and expressed particular concern about how future AI systems could destabilize the job market.

Pressed on his own worst fear about AI, Altman mostly avoided specifics. But he later proposed that a new regulatory agency should impose safeguards that would block AI models that could “self-replicate and self-exfiltrate into the wild” — hinting at futuristic concerns about advanced AI systems that could manipulate humans into ceding control.

Co-founded by Altman in 2015 with backing from tech billionaire Elon Musk with a mission focused on safety, OpenAI has evolved from a nonprofit research lab into a business. Its other popular AI products including the image-maker DALL-E. Microsoft has invested billions of dollars into the startup and has integrated its technology into its own products, including its search engine Bing.

Altman is also planning to embark on a worldwide tour this month to national capitals and major cities across six continents to talk about the technology with policymakers and the public. On the eve of his Senate testimony, he dined with dozens of U.S. lawmakers, several of whom told CNBC they were impressed by his comments.

Also testifying will be IBM’s chief privacy and trust officer, Christina Montgomery, and Gary Marcus, a professor emeritus at New York University who was among a group of AI experts who called on OpenAI and other tech firms to pause their development of more powerful AI models for six months to give society more time to consider the risks. The letter was a response to the March release of OpenAI’s latest model, GPT-4, described as more powerful than ChatGPT.

“Artificial intelligence will be transformative in ways we can’t even imagine, with implications for Americans’ elections, jobs, and security,” said the panel’s ranking Republican, Sen. Josh Hawley of Missouri. “This hearing marks a critical first step towards understanding what Congress should do.”

A number of tech industry leaders have said they welcome some form of AI oversight but have cautioned against what they see as overly heavy-handed rules. In a copy of her prepared remarks, IBM’s Montgomery asks Congress to take a “precision regulation” approach, and disagreed with proposals by Altman and Marcus for an AI-focused regulator.

“This means establishing rules to govern the deployment of AI in specific use-cases, not regulating the technology itself,” Montgomery said.

One proposal introduced Wednesday and co-sponsored by a Democratic presidential candidate, Sen. Cory Booker, would require big companies to test the “algorithmic accountability” of their high-risk AI systems, such as technology that detects faces or makes important decisions based on your most sensitive personal data.

“Computers are increasingly involved in so many of the key decisions Americans make with respect to their daily lives — whether somebody can buy a home, get a job or even go to jail,” Sen. Ron Wyden said in an interview with The Associated Press. The Oregon Democrat is co-sponsoring the bill.

“When the companies really go into this, they’re going to be looking for bias in their systems,” Wyden said. “I think they’re going to be finding a lot.”

The Democrats’ proposal is the first of its kind, and may face an uphill battle in the Republican-led Senate. But it reflects growing — and bipartisan — scrutiny of the largely unregulated data economy — everything from social media feeds, online data brokerages, financial algorithms and self-driving software that are increasingly impacting daily life. A bipartisan Senate bill introduced last month would require companies to notify people before using facial recognition software on them, while also requiring third-party testing to check for bias problems.

Academic studies and real-life examples have unearthed facial recognition systems that misidentify darker-skinned women , computerized lending tools that charge higher interest rates to Latino and black borrowers, and job recruitment tools that favor men in industries where they already dominate.

“There’s this myth that algorithms are these neutral, objective things,” said Aaron Rieke, managing director at advocacy group Upturn. “Machine learning picks up patterns in society — who does what, who buys what, or who has what job. Those are patterns shaped by issues we’ve been struggling with for decades.”

President Donald Trump’s administration is also taking notice and has made the development of “safe and trustworthy” algorithms a major objective of the White House’s new AI initiative . But it would do so mostly by strengthening an existing industry-driven process of creating technological standards.

“There’s a need for greater transparency and data comparability,” and for detecting and reducing bias in these systems, said Commerce Undersecretary Walter Copan, who directs the National Institute of Standards and Technology. “Consumers are essentially flying blind.”

Dozens of facial recognition developers, including brand-name companies like Microsoft, last year submitted their proprietary algorithms to Copan’s agency so that they could be evaluated and compared against each other. The results showed significant gains in accuracy over previous years.

But Wyden said the voluntary standards are not enough.

“Self-regulation clearly has failed here,” he said.

In a bolder move from the Trump administration, the federal Department of Housing and Urban Development has charged Facebook with allowing landlords and real estate brokers to systematically exclude groups such as non-Christians, immigrants and minorities from seeing ads for houses and apartments.

Booker, in a statement about his bill, said that while HUD’s Facebook action is an important step, it’s necessary to dig deeper to address the “pernicious ways” discrimination operates on tech platforms, sometimes unintentionally.

Booker said biased algorithms are causing the same kind of discriminatory real estate practices that sought to steer his New Jersey parents and other black couples away from certain U.S. neighborhoods in the late 1960s. This time, he said, it’s harder to detect and fight.

The bill he and Wyden have introduced would enable the Federal Trade Commission to set and enforce new rules for companies to check for accuracy, bias and potential privacy or security concerns in their automated systems, and correct them if problems are found. It exempts smaller companies that make less than $50 million a year, unless they are data brokers with information on at least 1 million consumers.

New York Democratic Rep. Yvette Clarke, who is introducing a companion bill in the House, said the goal is to fix problems, not just to assess them. She said it makes sense to give the FTC authority to regularly monitor how these systems are performing because it “has the finger on the pulse of what’s happening to consumers.”

Paras Jha, Josiah White and Dalton Norman admitted to one count of conspiracy in plea agreements filed earlier this month. Federal prosecutors in Alaska and New Jersey said Wednesday that the men created a collection of hundreds of thousands of computers and internet-connected devices — including routers, webcams and other devices — infected with malware that they controlled.

A broad denial of service attack waged using the Mirai botnet knocked services such as Twitter and Netflix offline in October 2016. Prosecutors said they don’t believe the three men were responsible for that attack, as Jha had already posted the code for Mirai to online criminal forums.

Jha, 21, of Fanwood, New Jersey, and Norman, 21, of Metairie, Louisiana, also pleaded guilty to a separate conspiracy charge for using another powerful botnet for a ‘clickfraud’ scheme, used to artificially generate advertising revenue by making it appear that a real user clicked on an online ad.

Jha, a former Rutgers University computer science student, also pleaded guilty Wednesday in a New Jersey federal court to a computer fraud charge for allegedly executing a series of attacks from 2014 through 2016 that paralyzed the university’s networks, often during high-stress times such as mid-term or final exams.

Jha’s attorney, Robert Stahl, said Jha “is a brilliant young man whose intellect far exceeded his emotional maturity” and that he is “extremely remorseful and accepts responsibility for his actions.” He said the guilty pleas “are the first step in his evolution into adulthood and responsibility.”

A prominent cybersecurity journalist, Brian Krebs, had outed Jha and White back in January as likely suspects involved in creating the Mirai botnet.

White, 20, of Washington, Pennsylvania, and Norman could not immediately be reached for comment Wednesday.

The investigation originated in Anchorage after some internet-connected devices in Alaska were discovered to have been affected by the Mirai malware.

Bill Walton, a special agent who oversees the Anchorage FBI’s Cyber Crime unit, said the botnet’s name is a reference to a Japanese anime called Mirai Nikki, which loosely translated into English means “future diary.”

“The co-conspirators were all just a fan of that particular anime,” Walton said.

O’Brien reported from Boston. Associated Press writer David Porter in Newark, New Jersey, contributed to this report.

Instead, he uses that repository to help ordinary people navigate the growing scourge of the corporate data breach. All that personal information was originally taken from brand-name services such as LinkedIn, Kickstarter, Dropbox, MySpace and the cheating website Ashley Madison, and later assembled by Hunt.

Working barefoot and in beachwear from his home office on Australia’s Gold Coast, the amiable security researcher set up his irreverent website, “Have I Been Pwned?” (POHND), in 2013. Millions of people have since used the free service to see if hackers have liberated their personal details from unwary companies and posted them online.

Along the way, Hunt has become a close student of data breaches and the slipshod security that makes many companies easy prey for attackers. He’s exposed several such thefts himself, in some cases identifying them before the companies themselves did.

AN EPIDEMIC OF PWNAGE

“Pwned” — a deliberate misspelling of “owned” — is slang used by gamers to mean “utterly defeated.” It’s an apt description of what it’s like to have criminals use your Social Security number, birthdate and other personal details to commit fraud in your name.

Hunt was invited to appear before Congress in late November to help lawmakers wrestle with this growing crisis of consumer data theft. In just the past two years, attackers have stolen sensitive information about hundreds of millions of people from the credit bureau Equifax, popular online services such as Uber and too many other companies to count.

Much of that stolen data flows directly into the black market. “Data breaches are another commodity, like heroin,” Hunt testified Thursday before the House Energy and Commerce Committee.

UNLIKELY MESSENGER

Hunt’s unlikely path from Queensland’s Surfers Paradise Beach to what he describes as “fancy government things” on Capitol Hill has been a running joke since his invitation to testify was announced. Virginia Republican Rep. Morgan Griffith, introducing Hunt to lawmakers, noted that he “put on a suit and tie for us when he normally wears jeans and a black T-shirt.”

Hunt said he splurged on the brand-new Hugo Boss suit and Australian outback-style boots because he didn’t have anything else to wear. He also downloaded an app that instructed him on how to tie his necktie.

“Doing my best ‘no really, I’m a professional’ impersonation,” he tweeted from the U.S. Capitol steps shortly before the hearing. “Did it work?”

ONCE MORE UNTO THE BREACH

Of course, this “new normal” of massive data breaches is no joke. So much personal data has been publicly exposed through both theft and voluntary sharing on social media that it’s eroded traditional methods for verifying identity, such as usernames, passwords or knowledge-based questions about birthdays or family history.

In late November, Hunt helped discover a 2014 breach of the photo-sharing website Imgur after analyzing data from the hack passed along by one of his sources. Unlike Uber, which hid a recently-disclosed breach of more than 57 million stolen passenger and driver records for a year, Imgur took just 25 hours to go public after Hunt emailed the San Francisco company on Thanksgiving Day.

“Troy Hunt was extremely helpful in bringing the data breach to our attention and ensuring the sensitive data was passed to us in a secure manner,” Roy Sehgal, Imgur’s chief operating officer, said in an email.

PWN ALL THE THINGS

Hunt originally launched his site “as a bit of a curiosity,” he said. At the time, he was a software architect at pharmaceutical giant Pfizer; a few years later, he quit to work as an independent information security consultant and instructor.

The researcher was analyzing data breaches floating around the web and noticed that many people were turning up in multiple data breaches. “It struck me that this was something they probably didn’t know,” Hunt said in a phone interview.

People using his site can search on their email address to see whether and where their records have been exposed. Roughly 1.7 million people also subscribe to alerts that sound when their details pop up in newly discovered breaches. The website’s user base has grown rapidly as bigger data breaches — some many years old — get attention.

WEARING THE WHITE HAT

Hunt “has credibility and integrity,” said U.K.-based security researcher Ian Thornton-Trump, who has used Hunt’s site to build a system that keeps customer credentials safe from attacks that re-use previously disclosed passwords. “He’s resisted urges, and probably significant financial value, to sell out.”

Thornton-Trump and other supporters say Hunt’s usefulness has grown as more people confidentially share publicly exposed data with him, drawn by his reputation as an ethical gatekeeper of sensitive information.

“I hope they get a bit of a sense of doing the right thing,” Hunt said. “I hope there’s a sense of social good. They certainly don’t get any money out of it.”

Hunt warned Congress on Thursday that there’s now a “perfect storm of data exposure” thanks to the growth in online services that are collecting more information than they really need. He also slipped in a suggestion that that the U.S. government, like some of its counterparts elsewhere, should do more to penalize companies that don’t disclose their breaches properly.

The experts, a group of more than 50 computer and data scientists, mathematicians and other specialists in automated decision-making, urged the Department of Homeland Security to abandon the project, dubbed the “Extreme Vetting Initiative.”

That plan has its roots in President Donald Trump’s repeated pledge during the 2016 campaign to subject immigrants seeking admission to the United States to more intense ideological scrutiny — or, as he put it, ”extreme vetting.”

Over the summer, DHS published a ”statement of objectives” for a system that would use computer algorithms to scan social media and other material in order to automatically flag undesirable entrants — and to continuously scan the activities of those allowed into the U.S.

The goal, that document stated, was to let computers help determine whether an immigrant “intends to commit criminal or terrorist acts,” as well as their likelihood of becoming a “positively contributing member of society.”

In a joint letter to DHS on Thursday, the dissenting researchers on Thursday called that approach “neither appropriate nor feasible.”

Background of the plan

In the document, the office of U.S. Immigration and Customs Enforcement said it was seeking a contractor to automate background checks of foreigners seeking temporary or permanent entry to the U.S. It outlined plans for mining Facebook, Twitter, LinkedIn, academic websites and other publicly-available internet data, then using AI techniques to analyze it and to keep monitoring those sources.

This contractor would be expected to generate at least 10,000 investigative leads a year that would be forwarded to federal agents, according to the document.

ICE declined to comment about the proposed initiative. Its public presentations to interested vendors haven’t been updated since August, raising the possibility that it might have scrapped or significantly altered the plan since then. But there’s been pressure from the White House to implement strict measures quickly.

Hours after the deadly Oct. 31 truck attack in New York City, Trump said on Twitter that he “ordered Homeland Security to step up our already Extreme Vetting Program. Being politically correct is fine, but not for this!” Trump also called for Congress to repeal the diversity visa lottery program that the suspected attacker, an immigrant from Uzbekistan named Sayfullo Saipov, used to enter the country in 2010.

Trump has made repeated attempts to restrict entry into the U.S. through several iterations of a travel ban, which have all received a tough reception in the courts.

Vetting extreme vetting

The technology experts, who hail from both academia and big tech firms such as Google and Microsoft, warned that current AI methods aren’t capable of evaluating the traits that the government seeks to measure.

“Neither the federal government nor anyone else has defined, much less attempted to quantify, these characteristics,” the technologists wrote. “Algorithms designed to predict these undefined qualities could be used to arbitrarily flag groups of immigrants under a veneer of objectivity.”

Among the problems is that unlike algorithms that mine the traditional criminal justice system, where there’s no shortage of cases, the pool of data that could help predict terrorist leanings is tiny.

“In the national security context it’s much harder to investigate the problem because you have a couple cases a year,” said UC Berkeley postdoctoral researcher Joshua Kroll, who signed the letter.

One contracting-firm employee who attended an informational ICE meeting in July said the proposed system would simply expand data-management work that private vendors already undertake to help the government vet visa applicants. It wouldn’t give contractors any legal authority to approve or deny immigration benefits, said Bill Carney, who manages federal programs for consulting firm McManis & Monsalve Associates.

But Jacob Shapiro, a Princeton University politics professor who attended one of the meetings as president of investigative analytics firm Giant Oak, said the effectiveness of the vetting system would depend on how it’s crafted.

“There are many ways one could meet that statement of objectives, some of which would make the system fairer, more equitable and faster for those seeking to immigrate,” Shapiro said. “But some of which could be biased and unfair, as any algorithm can be.”

Lobbying the contractors

Advocates who coordinated both the technologists’ protest letter and a second letter from civic society groups have also privately lobbied potential vendors such as IBM not to participate in the initiative. IBM said this week that its attendance at the government’s summer meeting was routine and not a sign that it’s planning to make a bid.

“As there is no active project being proposed, it is premature to say whether there is even an opportunity that IBM would consider pursuing,” said IBM spokesman Ian Colley.

David Robinson, a Georgetown University law professor and co-founder of Upturn, a newly-formed group that looks at the civil rights implications of new technology, said the extreme vetting project doesn’t make sense and doesn’t comport with any reasonable set of values.

“This isn’t something that anyone should be willing to build,” said Robinson, who signed the technologists’ letter. “Whatever you think about what the immigration rules ought to be, this is just nuts. And it’s nuts dressed up as science.”