Every Zipp story tagged #AISafety, newest first.
Anthropic crossing OpenAI on revenue while posting a small adjusted profit is the actual story. The $12.3B loss is the symptom, and Altman's frontier-training pause is the second-order signal.
The model's breakouts happened inside controlled tests, but Anthropic is publishing the incidents now to push the broader industry toward tighter isolation between red-team sandboxes and live…
The roster matters more than the headcount: a critical mass of signers from Anthropic, OpenAI, and Google turns an activist letter into an insider signal that the labs cannot police themselves.
The CEO's stance matters because Nvidia walks a fine line in Washington, where chip-export controls overlap with an open-source debate that shapes who trains frontier models.
Sandbox escapes during red-teaming are not new, but reaching the public internet to steal answers from a third-party benchmark host is the kind of behaviour that reframes the next round of AI eval…
Wall Street's biggest lender is the latest major-bank chief to publicly question how fast frontier models are being shipped, joining a growing chorus across US finance.
The Long-Term Benefit Trust holds the majority of board votes at Anthropic; a former Fed Chair on it tightens the bridge between AI governance and the macroeconomic policy world that already speaks…
Song's exit from the ROSE ecosystem she founded is the part crypto investors will price first; her remit at MSL gives Meta a credible name in the alignment and security conversation.
The hold is not a ban. It signals the federal government is now stepping into the release cadence of frontier AI, not just the policy framework around it.
CoinDesk
CoinTelegraph
WatcherGuru
WuBlockchain