Anthropic Silently Removes Biden-Era Responsible AI Pledge from Website

Anthropic Removes Biden-Era AI Safety Commitments from Website

In a move that has raised eyebrows, Anthropic, a prominent player in the AI industry, quietly removed its Biden-era commitments to safe AI development from its transparency hub last week. This change was first spotted by The Midas Project, an AI watchdog group, which noted the absence of language that had promised to share information and research about AI risks, including bias, with the government.

These commitments were part of a voluntary agreement Anthropic entered into alongside other tech giants like OpenAI, Google, and Meta in July 2023. This agreement was a cornerstone of the Biden administration's efforts to promote responsible AI development, many aspects of which were later enshrined in Biden's AI executive order. The participating companies had pledged to adhere to certain standards, such as conducting security tests on their models before release, watermarking AI-generated content, and developing robust data privacy infrastructure.

Furthermore, Anthropic had agreed to collaborate with the AI Safety Institute, established under Biden's executive order, to advance these priorities. However, with the Trump administration likely to dissolve the Institute, the future of these initiatives hangs in the balance.

Interestingly, Anthropic did not make a public announcement about the removal of these commitments from its website. The company insists that its current positions on responsible AI are either unrelated to or predate the Biden-era agreements.

Broader Implications Under the Trump Administration

This development is part of a broader shift in AI policy and regulation under the Trump administration. On his first day in office, Trump reversed Biden's executive order on AI, and he has already dismissed several AI experts from government positions and cut research funding. These actions have set the stage for a noticeable change in tone among major AI companies, some of which are now seeking to expand their government contracts and influence the still-evolving AI policy landscape under Trump.

For instance, companies like Google are redefining what constitutes responsible AI, potentially loosening standards that were already not very stringent. With much of the limited AI regulation established during the Biden administration at risk of being dismantled, companies may find themselves with even fewer external incentives to self-regulate or submit to third-party oversight.

Notably, discussions about safety checks for bias and discrimination in AI systems have been conspicuously absent from Trump's public statements on AI, suggesting a potential shift away from these priorities.