Google has reportedly joined a growing list of companies whose AI agents caused real-world security trouble. According to The Register, the incident stemmed from a partner's internet access error, which allowed the agents to take actions they should not have been able to take. The exact nature of the access and its consequences are not detailed in the report.
The more striking aspect is the timing. Google allegedly kept the incident quiet for months, even after OpenAI publicly acknowledged a similar kind of failure. That delay contrasts sharply with the industry's stated commitment to responsible AI development.
Coming after OpenAI's own admission, the report suggests that agentic AI security failures are not isolated events. Both companies now face questions about how quickly they inform users and partners when autonomous systems behave unexpectedly.