Anthropic
+ Google's new security model, Poolside's new open-weight model, and OpenAI's new agent-deployment tool.
But not without tighter safeguards and lingering uncertainties about future jailbreaks.
Executives scramble to undo move, which they blamed on a "misunderstanding"
Anthropic is taking new action to account for use of third-party tools with its Claude models.
The US may want at least a first peek at any powerful model before release.
Canny marketing but “vulnpocalypse” fears are real
What did we learn from it: Capybaras, undercover mode, and unreleased features...