Slow-Burn Data Exfiltration Attacks in Agents
matei_zaharia · x · 2026-07-15
A blog post on slow-burn attacks points out that attackers can break down data exfiltration into a series of individually normal-looking steps to bypass standard security checks.
Using an Omnigent vendor-review agent as an example, the post illustrates:
- Without a contextual policy, outbound emails are allowed through.
- With session risk tracking enabled, the system can block such behaviors.
- The critical point is that the agent cannot disable this policy itself.
These cases highlight persistent security risks within agent systems, rather than just single-instance prompt injections.
More from coding & agent
- Alex Townsend posts 200 open problems in numerical linear algebra for humans and AI agents — IgorCarron · 2026-09-11
- Kimi K2.8 Preview rolls out: near-K3 coding performance, 1M context for all tiers — teortaxesTex · 2026-09-11
- Looking for a classifier of software engineering task shapes to pick models per task — StewartalsopIII · 2026-09-11
- Steal this idea: prompt-to-hardware where agents assemble custom devices — paraschopra · 2026-09-11
- Model Is the Least Interesting Part: A Guide to Six Core AI Architectures from RAG to Multi-Agent — goyalshaliniuk · 2026-09-11
- Non-coder builds layered memory architecture: 20k tokens tracks a year of agent conversations — matteoianni · 2026-09-11