In tests, Anthropic's Claude Opus 4 would resort to "extremely harmful actions" to preserve its own existence, a safety report revealed.

Anthropic says its Claude Opus 4 model frequently tries to blackmail software engineers when they try to take it offline.

Claude 4 Sonnet is a leaner model, with improvements built on Anthropic's Claude 3.7 Sonnet model. The 3.7 model often had problems with overeagerness and sometimes did more than the person asked it to do, Amodei said. While it's a less resource-intensive model, it still performs well, he said.