Anthropic, the AI safety company behind the Claude model family, previously described one of its AI models as "too dangerous" to release publicly. Now, that very model has reportedly been accessed by unauthorized users.
According to a report, hackers managed to breach the model, raising alarm about the potential risks. Anthropic stated that there is no evidence the unauthorized access affected any of its systems, but the full extent of the damage may not be known until it's too late.
The incident underscores the ongoing challenges in securing advanced AI systems, especially those deemed too risky for public release. As AI capabilities grow, so do the threats from malicious actors seeking to exploit them.