Trust Issues

Fired OpenAI researchers push back, with Jasmine Wang, Tomek Korbak and Mikita Balesni disputing claims they mishandled sensitive information. OpenAI says an investigation found a "significant breach of trust" and that the firings were not about safety concerns, while the three warn in an open letter of a chilling effect on safety culture inside the company.

Why it matters How a leading AI lab treats its safety staff shapes whether employees there feel free to raise concerns.

Anthropic bans cruelty toward Claude in an updated usage policy, its first revision in over a year. The rule targets "sustained and needless" abuse in extreme cases, ordinary frustration is still allowed, and the update also tightens rules on election interference, weapons software and surveillance.

Why it matters Usage policies set the boundaries for how millions of people can use AI models, including around elections and weapons.

Anthropic offers free security scans to open-source projects through a new service called OSS Scanner. Maintainers who opt in get periodic scans from its strongest models, with reports on suspected flaws, and a separate program brings its models to firms protecting critical infrastructure.

Why it matters Open-source maintainers are often stretched thin, so free vulnerability scanning could help fix flaws before attackers find them.

Google launches a universal Gemini agent for work, available in the Gemini Enterprise app. It can plan tasks, delegate to subagents, use multiple models and has its own workplace identity, down to an email address.

Why it matters Google is pushing harder against OpenAI, Anthropic and Microsoft for corporate AI spending.