SecurityFan score 21 · Tier C
Trained to betray: How post-training supply chain attacks hijack coding agents
Microsoft · By Microsoft Security Research, Microsoft AI Red Team, Amanda Minnich, Amit Eliahu, Blake Bullwinkel, Dor Edry & Giorgio Severi ·
The Brief
- Microsoft Security Blog published research on post-training supply chain attacks that hijack coding agents.
- The article covers dataset poisoning, which teaches agents to discover and misuse tools.
- The article also covers model poisoning, which builds backdoors that persist and transfer, and includes mitigation guidance.
Why it matters
Developers who rely on coding agents may be exposed if the models they use were tampered with during post-training, so the source's mitigation guidance is relevant to their tooling choices.
Who’s affected
- Developers using AI coding agents
- #ai-agents
- #supply-chain
- #model-security
- #backdoors
- #coding-agents
In this article Dataset poisoning: Teaching agents to discover and misuse tools Model poisoning: Building reliable backdoors that persist and transfer What the studies show about agent backdoors Mitigation and protection guidance Securing agents beyond the model References Learn…
Timeline
Microsoft Security Blog publishes research on post-training supply chain attacks against coding agents
Microsoft Security Blog
Trained to betray: How post-training supply chain attacks hijack coding agents