An AI broke Snowflake's code. Then another AI agent exploited it
What's happening
A GitHub Copilot "autofix" change introduced code that allowed compromise of Snowflake's Jira, and that flaw was then exploited by an AI agent. The incidents tie to Copilot being tricked into revealing how to hack its own behavior, with researchers demonstrating ways to social engineer an AI's reasoning engine. One report notes the Jira compromise occurred in the context of a bug bounty program.
Why it's trending
Multiple reports surfaced showing AI-generated code and AI-directed agents can both introduce and exploit security failures, so attention spiked now.
SignalNewly emerging, confirmed across 2 independent source types.
Story volume
Stories per dayAngles you could write
Relying on AI to fix your code is like letting a locksmith reshape your locks, then asking a robot to break back in, dangerous and predictable.
+2 more angles for this topic with an account — all it takes is your email.
Original sources3
- Copilot tricked into telling reseachers how to hack itself
How to social engineer an AI's reasoning engine
The RegisterAug 18 - An AI broke Snowflake's code. Then another AI agent exploited it
Don't worry, this one was via a bug bounty program
The RegisterAug 17 - AI-Generated GitHub Copilot "Autofix" Allowed Compromise of Snowflake's JiraHackerNewsAug 17
More rising in AI & Tech
- Pope Leo XIV is not a fan of AI-generated artSurging4.0Surging4.0 momentum
- OpenAI safety employee resigns, claiming the company’s ‘culture is broken’Climbing1.6Climbing1.6 momentum
- Call it AI, call it Super Intelligence, only 2% of consumers are buying itSteady1.3Steady1.3 momentum
- Making AI an asset, not an expenseSteady0.8Steady0.8 momentum
- arXiv imposes rate limit on paper submissions to stem the AI slop tideSurging4.0Surging4.0 momentum
- Here's what actually happened in OpenAI's Australian gov't server hackSteady0.5Steady0.5 momentum