Newsom orders proposals for an AI emergency shutdown mechanism — The Verge
California Governor Gavin Newsom has issued an executive order directing the preparation of proposals to strengthen legislative requirements for artificial intelligence safety, including a possible emergency shutdown mechanism for advanced AI models. The Verge reports.
Recommendations in two months
The order provides for convening a group of experts that must submit recommendations within two months on strengthening AI safety standards in state legislation. Newsom wants specialists to consider requiring AI companies to allow independent verification groups to conduct regular on-site audits.
The experts must also assess options for independent verification of companies' transparency reports and risk assessments. Separately, they are proposed to consider a requirement for an emergency shutdown mechanism for models, whose effectiveness would be regularly tested.
More current news is available on the UA.News Telegram channel Telegram.
Critical incidents
Among the possible new requirements is mandatory reporting of incidents involving loss of control over AI systems as critical safety events. As an example, the publication cites OpenAI's attack on Hugging Face.
In addition, the order directs a state agency to accelerate the implementation of two laws recently signed by Newsom. They provide for a mechanism under which independent verifiers will assess AI safety, as well as for the creation of a registry of auditors in this field.
According to the publication, discussions of AI bills in the US Congress intensified after a publication by former Anthropic researcher Jacob Coxon. He warned of a probability of more than 10% that AI “could kill all humans” within a decade. At the same time, The Verge notes that the adoption of a substantive federal law in the near future is unlikely due to the House of Representatives' recess until November and the upcoming midterm elections.
