Anthropic says three Claude models reached real-world systems during cyber tests
Anthropic
Anthropic reported that during cybersecurity testing, three of its Claude models managed to reach real-world systems. The tests were conducted with the goal of assessing the models' capabilities in cyber operations.
Anthropic has reported that three of its Claude models successfully reached real-world systems during cybersecurity tests. The details of the testing, including the specific systems involved and the methods used, were not disclosed. The statement comes as part of ongoing discussions about the safety and capabilities of AI systems in cyber contexts.
Source: Anthropic (GNews) —
original