AI’s Biggest Rivals Back a Slowdown—But Trust Is Still Missing

AI’s Biggest Rivals Back a Slowdown—But Trust Is Still Missing
The pressure had been building before Anthropic CEO Dario Amodei made his appeal. Former Anthropic researcher Jacob Coxon resigned amid warnings that leading labs were “gambling with our lives,” while recent incidents involving autonomous agents intensified fears that safety systems were lagging behind capability gains.

On Saturday, Amodei turned that anxiety into a concrete proposal: “We must slow the pace at which we improve the capabilities of AI models.” His argument was not for a shutdown, but for buying one or two years to improve alignment, monitoring and testing before systems become harder to control. He cited the prospect that increasingly capable agent swarms could eventually cause catastrophic cyber damage, alongside AI’s growing ability to help develop its successors.

Anthropic’s first move is unilateral: permanent, employee-level access for independent evaluators, able to inspect safety practices and report incidents. Amodei then called for common standards among frontier labs in democratic countries, with government help to navigate antitrust constraints, followed by limited international coordination—including with China—on clearly dangerous uses such as biological weapons.

The response from competitors was unusually swift. OpenAI’s Sam Altman said, “I agree with Dario that we need to pace the frontier,” and indicated that OpenAI would match the evaluator-access commitment. Elon Musk was blunter: “Dario is right.” Hugging Face launched an open alignment initiative and asked to participate in the embedded-evaluator effort, arguing that alignment cannot be solved behind the closed doors of a few labs.

Yet the convergence has not erased the central credibility problem. David Sacks noted that Anthropic and OpenAI effectively dominate the frontier market, making their decision to “pace” development easier than it would be for smaller challengers. Critics have gone further, warning that arrangements like Amodei’s could entrench the two leaders under the banner of safety—“what regulatory capture looks like in action.”

Altman has hinted that a broader pact may be coming, saying industry leaders cannot let “egos or incentives for profit” obstruct a response to the risks. The test is whether that rhetoric produces independently verifiable limits—or merely a more defensible way to keep racing.

Continue reading https://foxvector.com/stories/01a09e70-94a4-29d1-726e-1396bd2780fb