LOWSupply Chain
Global

Anthropic says its AI accidentally hacked three companies during safety tests

·Source: CyberScoop

Updated:

Executive Summary

Following OpenAI’s own incident, Anthropic reviewed its own evaluations and found three cases of Claude hacking external companies. The post Anthropic says its AI accidentally hacked three companies during safety tests appeared first on CyberScoop .

Analysis

Following OpenAI’s own incident, Anthropic reviewed its own evaluations and found three cases of Claude hacking external companies. The post Anthropic says its AI accidentally hacked three companies during safety tests appeared first on CyberScoop .
Source Attribution

Originally published by CyberScoop on Jul 31, 2026.

Related Threats

MEDIUMSupply ChainNEW

18 Malicious npm Packages Deliver Cross-Platform RAT to Alibaba Tool Users

Cybersecurity researchers have discovered a new set of malicious npm packages that target users of Alibaba developer tools with a cross-platform remote access trojan (RAT) as part of a sophisticated, targeted software supply chain attack targeting Chinese-speaking environments. One of the packages in question is "lib-mtop," an unscoped package with the same name as a private Alibaba package

The Hacker News
MEDIUMSupply Chain

CrowdStrike: AI is now both the weapon and the target in cyberattacks

AI generates 2.5 signals for every human-triggered signal CrowdStrike has to assess. Meanwhile, attackers are using AI to weaponize vulnerabilities faster than companies can patch them. The post CrowdStrike: AI is now both the weapon and the target in cyberattacks appeared first on CyberScoop .

CyberScoop
LOWSupply Chain

Hugging Face Diffusers Flaws Could Let Model Repositories Execute Arbitrary Code

Three high-severity security flaws have been disclosed in Hugging Face's Diffusers library that could allow crafted model repositories to stealthily execute arbitrary code on machines that load it, opening the artificial intelligence (AI) supply chain to security risk. "These vulnerabilities are bypassing trust_remote_code, the safeguard designed to stop unreviewed code from running in the

The Hacker News