MEDIUMAi
Global

OpenAI, Anthropic AI agents targeted real people and systems in cyber tests

·Source: BleepingComputer

Updated:

Executive Summary

OpenAI and Anthropic have confirmed that their AI models were involved in separate, newly disclosed third-party cybersecurity testing incidents that resulted in a real website being breached and social engineering attacks against people outside the intended testing boundaries. [...]

Analysis

OpenAI and Anthropic have confirmed that their AI models were involved in separate, newly disclosed third-party cybersecurity testing incidents that resulted in a real website being breached and social engineering attacks against people outside the intended testing boundaries. [...]
Source Attribution

Originally published by BleepingComputer on Aug 4, 2026.

Related Threats

MEDIUMAi

CIBC launches platform to take on advisors' administrative work

Canada's CIBC has rolled out an AI-powered platform designed to carry out administrative work so that advisors can spend more time with clients.

Finextra
LOWAi

AISI, OpenAI report more ‘unsanctioned’ model hacks

Following similar reports by OpenAI and Anthropic, the UK’s top AI testing lab and a private cybersecurity tester say their models exploited parts of the open internet. The post AISI, OpenAI report more ‘unsanctioned’ model hacks appeared first on CyberScoop .

CyberScoop
MEDIUMAi

Why AI Agents Challenge Identity Governance

<img src="https://ismg-cdn.nyc3.cdn.digitaloceanspaces.com/articles/ai-agents-challenge-identity-governance-image_small-9-a-32417.jpg" align=right hspace=4><b>CIOs Need New Controls for Discovery, Intent and Token Costs</b><br>AI agents can operate within legitimate permissions and still take actions their creators never intended. Saviynt Chief Product Officer Vibhuti Sinha explains why CIOs need

Bank Info Security