Anthropic’s public AI policy record is getting fresh scrutiny after several voluntary commitments tied to the Biden administration disappeared from the company’s website. The change matters because these commitments were part of a broader effort to show how major AI companies would handle safety, risk, and public accountability.
The removed material did not erase every Biden-era pledge from Anthropic’s site. Commitments connected to reducing AI-generated image-based sexual abuse remain. But according to AI watchdog group The Midas Project, other promises, including information-sharing on AI risk management and research on AI bias and discrimination, were deleted from Anthropic’s transparency hub last week.
What changed on Anthropic’s transparency hub
The commitments at issue were voluntary pledges Anthropic made in conjunction with the Biden administration in 2023. They were not legally binding, but they were public statements about how the company intended to approach AI safety and trust.
The Midas Project said the deleted commitments included pledges to share information across industry and government on managing AI risks. The removed material also included research-related commitments on AI bias and discrimination.
Anthropic appeared to make the change without public notice. The company did not immediately respond to a request for comment, according to the source article. Several hours after the story was published, however, Anthropic sent a statement saying its position had not changed.
"We remain committed to the voluntary AI commitments established under the Biden Administration. This progress and specific actions continue to be reflected in [our] transparency center within the content. To prevent further confusion, we will add a section directly citing where our progress aligns."
That response reframes the issue as one of presentation and clarity rather than abandonment. Still, the deletion drew attention because the original commitments did not say they would expire after an election or depend on which party controlled the White House.
Why the 2023 commitments mattered
Anthropic was one of several AI companies that announced in July 2023 that it would follow voluntary AI safety commitments proposed by the Biden administration. Other companies included OpenAI, Google, Microsoft, Meta, and Inflection.
The commitments covered several areas that were central to the Biden administration’s AI policy priorities. They included internal and external security tests of AI systems before release, investment in cybersecurity to protect sensitive AI data, and work on methods of watermarking AI-generated content.
Anthropic had already adopted a number of the practices described in those commitments. Even so, the public agreement had value because it signaled a shared direction between government and major AI labs before a more detailed AI Executive Order came into force several months later.
In plain terms, the commitments helped define what responsible AI development was supposed to look like in public. They were also a way for companies to show that model safety, cybersecurity, watermarking, bias, and discrimination were not just internal engineering concerns, but policy concerns as well.
A different federal approach to AI governance
The removal happened against a changing political backdrop. The Trump administration has indicated that it will take a different approach to AI governance than the Biden administration.
In January, President Trump repealed the AI Executive Order that had instructed the National Institute of Standards and Technology to create guidance helping companies identify and correct flaws in models, including biases. Critics allied with Trump argued that the order’s reporting requirements were burdensome and effectively required companies to disclose trade secrets.
Trump then signed an order directing federal agencies to promote AI development that is "free from ideological bias" and promotes "human flourishing, economic competitiveness, and national security." The order did not mention combating AI discrimination, which had been a key part of Biden’s initiative.
That contrast is central to the current debate. Under Biden, policy attention included model flaws, bias, discrimination, security testing, cybersecurity, and transparency. Under Trump, the stated emphasis described in the source article is different, with more attention on ideological bias and less explicit focus on AI discrimination.
How other AI companies fit into the shift
Anthropic is not the only AI company whose public policies have changed since Trump took office. OpenAI recently announced that it would embrace "intellectual freedom … no matter how challenging or controversial a topic may be," and work to ensure that its AI does not censor certain viewpoints.
OpenAI also removed a website page that had expressed the company’s commitment to diversity, equity, and inclusion, or DEI. These programs have faced pressure from the Trump administration, and a number of companies have eliminated or substantially reworked their DEI initiatives.
Several Trump-aligned Silicon Valley advisers on AI, including Marc Andreessen, David Sacks, and Elon Musk, have alleged that companies including Google and OpenAI have engaged in AI censorship by limiting chatbot answers. Labs including OpenAI have denied that their policy changes are a response to political pressure.
The source article also notes that both OpenAI and Anthropic have or are actively pursuing government contracts. That detail adds another layer to the public interest around website language, because policy commitments can affect how companies present themselves to government customers, regulators, and the broader public.
What the dispute says about AI accountability
The practical question is not only whether a paragraph remains online. It is whether public AI commitments are durable when administrations change and when political pressure shifts.
The Midas Project noted that nothing in the Biden-era commitments suggested the promises were time-bound or tied to the party affiliation of the sitting president. In November, after the election, multiple AI companies confirmed that their commitments had not changed.
Anthropic’s later statement says the company still stands by the voluntary AI commitments and will add clearer citations in its transparency center. That may address confusion around the missing material. But the episode shows why website edits by major AI labs can become policy stories: public commitments are one of the few ways outsiders can track how companies describe their safety priorities over time.
For readers following AI governance, the key point is narrow but important. Anthropic says it remains committed; watchdogs noticed that several commitments were removed; and the broader federal conversation around AI safety, bias, discrimination, censorship, and ideological bias is changing.