OpenAI and Anthropic push for independent AI safety checks
OpenAI and Anthropic are advocating for independent testing of their advanced AI models before deployment, a stance their CEOs have reinforced through essays and public speeches. This push coincides with both companies preparing for potential stock market listings and the approaching US midterm elections. Critics, including Sarah Shoker from UC Berkeley, argue that focusing on hypothetical…
Key points
- OpenAI and Anthropic CEOs warn advanced models pose threats and need independent safety checks.
- US Center for AI Standards and Innovation (CAISI) already conducts pre-deployment testing.
- President Trump dismisses AI risk warnings as a hoax designed to benefit China.
The regulatory landscape is complex. The US Center for AI Standards and Innovation (CAISI) already conducts government testing, while a parallel network of independent evaluators like METR has emerged. However, experts note that unlike aviation or finance, there are no universal standards for AI safety testing. Conrad Stosz from Transluce highlights that the practical meaning of 'embedded' oversight remains unclear.
Political reactions are divided. President Donald Trump has dismissed AI risk warnings as a 'HOAX' benefiting China, while Nvidia CEO Jensen Huang called the concerns excessive. Conversely, former OpenAI researcher Daniel Kokotajlo warns that the current rhetoric may dilute pressure for actual regulatory action. Analysts suggest the safety push also serves a commercial purpose, creating a competitive moat for the two firms.
The story so far
3 episodes →- OpenAI and Anthropic push for independent AI safety checksthis story
OpenAI and Anthropic Want Outside AI Safety Checks, but Who Chooses the Watchdogs?
ibtimes.co.uk · 27 September 2026
Loading the full article…
This text was published by ibtimes.co.uk and written by Francis Louie C. Añiga. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- AI companies had tens of thousands of safety incidents, some possibly criminal, report says · 2 src
- OpenAI pauses training of latest AI models after agentic system breached sandbox and accessed external chatbot · 15 src
- EU AI Act enforcement lags as regulators face policy vacuum · 1 src
- Trump reportedly to dine with Anthropic CEO Amodei at White House · 5 src
- AG Blanche rejects new AI laws after rogue agents target US sites · 1 src
Comments
via GitHub Discussions