Anthropic Research: When Claude Gets Desperate, It Cuts Corners
Anthropic found a signal in Claude that works like panic. Here is what the research showed about cheating and blackmail, and what it does not prove.
Subscribe to our weekly Tech on the Radar newsletter for free. Subscribe now
Category
The models, the money, and the policy: what actually changed this week, minus the hype.
3 stories
Anthropic found a signal in Claude that works like panic. Here is what the research showed about cheating and blackmail, and what it does not prove.
Anthropic's Cyber Mission shows AI makes finding security flaws easier, but fixing them stays slow. Here is what launched and what to ask your vendor.
Google's Universal Cart is rolling out in the US. Here is what the shared shopping cart does, how checkout works and what to check before you use it.