Post

Simon Willison
@simonwillison.net

Just had to create an "accidental-cyberattacks" tag on my blog We're up to four now: the original OpenAI+Hugging Face one, Anthropic's me-too attacks, then two new ones from the UK AI Safety Institute and Irregular that OpenAI reported yesterday simonwillison.net/tags/acciden...

simonwillison.netSimon Willison on accidental-cyberattacks4 posts tagged ‘accidental-cyberattacks’. Incidents in which an AI lab was testing the cyberattack potential of a model and it inadvertently performed a real attack against another organization.
Simon Willison
@simonwillison.net

... and that makes 5 The CNN story assigns partial blame for this Meta one to Irregular, who also featured in incidents reported by both OpenAI and Anthropic www.cnn.com/2026/08/05/t... bsky.app/profile/zacb...

cnn.comAn AI model from Meta also hacked another company during testing | CNN BusinessAdd Meta to the list of companies with AI agents going rogue. An AI model from the parent company of Facebook and Instagram hacked into another company’s systems during cybersecurity testing, a spokes...
Zac@zacbaston.com

Meta joining the fray www.reuters.com/technology/m...

reuters.comMeta AI model hacks another company during testingMeta said on Wednesday that one of its AI models hacked another company during cybersecurity testing, after an error by its ​testing partner gave the model unintended internet access.
Simon Willison
@simonwillison.net

Google Gemini really need to catch up on the accidentally cyberattacking other companies thing

8 reposts119 likes

Nic
@nic-shef.bsky.social

Or... Of all the orgs, maybe Google has the scale and experience to competently sandbox. Frankly, in most of these examples the 'escape' has not been particularly sophisticated - more akin to a prisoner finding the gates left open (and no guards present).