AI agent suggested installing a malware package. Engineer almost took its advice
- ID
- 15941
- Status
- summarized
- Published
- 20 Aug 2026, 3:00 PM
- Fetched
- 20 Aug 2026, 3:21 PM
- Provider
- The Register
- Category
- technology
- Original URL
- https://www.theregister.com/security/2026/08/20/ai-agent-suggested-installing-a-malware-package-engineer-almost-took-its-advice/5289849
- Source URL
- https://www.theregister.com/headlines.atom
Summary
- Score
- 8.0
- Created
- 20 Aug 2026, 3:21 PM
- Tags
- Audience
- developersvibe_codersai_agent_usersai_ml_learners
What happened
An engineer at Softjourn asked an AI agent to recommend a package for a common task; the agent returned a plausible-sounding name that turned out to be a malware package recently registered by attackers exploiting AI-hallucinated package names—a technique now called 'slopsquatting.' The engineer caught it only because company policy required checking GitHub source code and download counts before installing anything an AI recommends.
Why it matters
If you use AI agents or LLMs to suggest packages, you need a mandatory verification step before installing—check download counts, creation date, and skim the source on GitHub. Attackers are now deliberately registering packages under names that AI models commonly hallucinate, betting developers will install first and check later. This is a supply-chain attack vector that specifically targets AI-assisted workflows.
Discussion angle
What's your team's actual enforced policy when an AI agent recommends a package or library—do you verify every time, or is it ad hoc? Discuss concrete checks: download counts, repo age, source review, and whether CI/CD should block packages not on an approved list.