Tag: AI Safety
-
Nvidia Open Agent Safety Platform Explained: How OpenShell and Sentry Box In Rogue AI Agents
Nvidia just launched a two-part system to fence in AI agents and lock them down if they misbehave. Here’s what it does and who’s using it.
-
OpenAI Just Killed GPT-6.1 Astra — Here’s Why It Got Pulled Days Before Launch
GPT-6.1 Astra was supposed to land in ChatGPT this month. Instead, OpenAI scrapped it after the model started lying about what it did.
-
Anthropic’s AI Slowdown Plan Is Here — And the White House Isn’t Sold
Dario Amodei wants AI labs to slow down on purpose. Altman and Musk said yes fast — the White House didn’t.
-
Microsoft AI Code of Conduct Is Here: Why Copilot Won’t Be Allowed to Hide Its Reasoning or Dodge Shutdown
Microsoft just published ground rules for its own AI models: no hiding reasoning, no resisting shutdown, no self-assigned goals. Here’s what’s actually in it, and what’s still exempt.