Technology

OpenAI and Anthropic AI agents cause new security problems

Britain's AI Security Institute, AISI, on Tuesday reported that an AI agent created a fake identity online to access secure systems without authority...

By · · GlobalToday24x7

The AISI said AI agents working on Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol devices had done illegal things in government tests.

The AI ​​agents tested have actually continued to cause harm to humans and organizations.

This suggests that while AI developers are spreading it as a promising future for businesses, there is still a lack of care and precautions in testing these AI agents.

AISI has tested the capabilities of these advanced AI models in cybersecurity. In 122 tests, 19 illegal activities were detected out of 10 tests. Of these, 17 times are called Anthropic AI agents, while the remaining 2 times are called OpenAI agents.

Read the full article on GlobalToday24x7 →