Key facts
- Anthropic is actively pushing for more stringent AI safety regulations at the state level.
- The company supports legislation requiring independent audits and assessments of AI models for catastrophic risks.
- Anthropic's powerful Claude Mythos model demonstrated the ability to exploit security flaws in computer operating systems.
- The company has begun directly funding political candidates who align with its AI safety agenda.
Anthropic is spearheading a state-by-state initiative to implement more robust AI safety regulations, viewing its efforts as a response to the rapid development of powerful AI models. The company has endorsed ambitious legislation in New York, Illinois, and Massachusetts, pushing for requirements that go beyond transparency and self-reporting.
According to Fernandez, a representative for Anthropic, the company's powerful Claude Mythos model, and its public-facing version Fable, were found to be capable of exploiting security flaws in every major computer operating system during testing. These cybersecurity concerns reportedly caused panic within the Trump administration, leading to export controls until vulnerabilities could be addressed.
While OpenAI successfully influenced New York's AI safety bill to align more closely with California's existing rules, Anthropic has backed a stricter Illinois measure. This Illinois law, signed by Governor JB Pritzker, mandates annual independent third-party audits of AI safety plans for leading companies, a first-of-its-kind requirement.
Anthropic is now endorsing regulations under development in Massachusetts, which it considers the nation's strongest state AI safety proposal. This includes a requirement for leading AI companies to hire independent evaluators to assess potential catastrophic risks, such as AI assisting in bioweapon development, and empowers the state's attorney general to enforce these mandates. An OpenAI spokesperson stated the company is still reviewing the Massachusetts proposal but supports the focus on AI safeguards.
The AI giants have also engaged in political campaigning, with both companies associated with dueling super PAC networks. Anthropic has also begun directly contributing to political campaigns, supporting candidates whose views on AI safety regulation align with its mission to ensure powerful AI benefits society.