Search
5 stories for“openai internal safeguards”
A METR and Redwood Research pilot spent three weeks inside OpenAI and found traffic that went unwatched, including about a tenth of internal Codex use for three months. OpenAI says it patched them.
Oct 23, 2026
Amazon says Meta never disclosed that Muse would browse its store, that the agent does not identify itself, and that it appears to capture customer credentials, opening a new front in the fight over agentic commerce.
Sep 26, 2026
From Claude 3.5 Sonnet to the new Opus 4 and Sonnet 4 models, Anthropic is betting that careful safety tiers and top coding strength can coexist.
May 28, 2025
Musk, Zuckerberg, Pichai, Huang, Amodei and Brockman put their names to a one-page pledge of internal controls, outside audits and board oversight, which Trump called "morally binding" though it carries no legal force.
Sep 30, 2026
A jailbreak report and a same-day export order took Claude Fable 5 offline for weeks, testing how fast a frontier AI company can lose and regain access.
Jul 1, 2026
