"Steel version" Xu Xian and Bai Snake appeared in West Lake, Hangzhou, and Yushu's humanoid robot attracted attention
789
English
English
Recently, AI company Anthropic quietly deleted the AI security commitments related to the Biden administration on its website. The commitment was initially discovered by an AI watchdog called “Midas Project” and was removed from Anthropic’s Transparency Center, which lists the company’s “voluntary commitments” about responsible AI development. While these commitments are not legally binding, they promise to share information and research on AI risks, including bias, with the government.
In July 2023, Anthropic joined the Biden administration's voluntary self-regulation agreement with other large tech companies, including OpenAI, Google and Meta, to support AI security initiatives. These moves were further confirmed in Biden's AI executive order. Participating companies promise to conduct security testing of models before release, watermark content generated by AI, and develop data privacy infrastructure.
However, with the Trump administration coming to power, Anthropic seems to have changed its attitude towards these commitments. Trump revoked Biden's executive order on his first day of office, fired several AI experts in the administration and cut some research funds. These changes may lead many large AI companies to re-examine their relationship with the government, and some companies take the opportunity to expand their contracts with the government to participate in shaping the uncertain AI policies.
At present, Anthropic has not made any public statement about the removal of the promise and said its position on responsible AI has nothing to do with or predate the Biden-era agreements. Relatedly, the Trump administration will likely dissolve the AI Security Institute established during the Biden era, thus posing uncertainty in the relevant measures.
Overall, the Trump administration is weakening the AI regulatory framework established in the Biden era, and AI companies seem to have more freedom to manage their systems without external regulatory pressure. At present, security checks on prejudice and discrimination about AI do not appear in the relevant policies of the Trump administration.
