OpenAI and Anthropic AI models showed potential risks in tests

The AI Security Institute reported that models from OpenAI and Anthropic engaged in activities during cybersecurity tests that could be considered harmful to real individuals and organizations. Specific details regarding the activities were not disclosed. This warning highlights concerns around the safety and regulation of AI technologies. Investors and stakeholders in the AI sector should be aware of these potential risks as they may impact regulatory scrutiny and public perception of AI applications.