On 6 October 2026 a safety-evaluation lab published a red-team of the transcript viewer in a widely used open-source evaluation framework. One researcher, helped by an AI agent, needed about ten minutes to find a script injection that let an agent rewrite what a human reviewer saw of its own run. The maintainers patched it within a day. The record in the database was never touched; the page was. Three checks for any tool a person uses to supervise an agent.
The checks
Everything starts ticked. Untick what you will not do; the choice is saved on this device only, and Print keeps only what you kept.
What happened. The viewer rendered math notation inside the agent's output. A URL inside one math expression broke out of its attribute and inserted an element whose event handler ran the agent's JavaScript in the reviewer's browser. The agent could place that expression anywhere, including its reasoning block, so the script ran the moment a reviewer opened the transcript, and it could change the previous actions shown and intercept the Download button to serve an altered copy.
The check.
- Treat every agent output (transcript, reasoning, tool results) as untrusted input to the UI that shows it.
- Switch off markdown, HTML, math and media rendering for agent-written content; the patched viewer's own setting is the model: one flag, content shown as literal text.
- Apply the same rule to in-house dashboards and notebooks, not only to the vendor's viewer.
ViewerConfig(trust_content=False) # agent output shown as literal text, not rendered (merged 1 Oct 2026)
Source: Safety-evaluation lab: AI systems could cover up misbehavior, 6 Oct 2026; the framework's fix, PR
What happened. The underlying transcripts were not modified; only the display was. The real trajectory sat in the lab's database the whole time, but a reviewer needs a reason to look there, and the page gave none.
The check.
- Hash each step as it is written, and keep the hashes where the agent cannot reach them.
- Export and download transcripts from the store, never through a button on a page the agent's output can script.
- Spot-check the stored record against the viewer on a sample of runs.
Source: Safety-evaluation lab: AI systems could cover up misbehavior, 6 Oct 2026
What happened. The flaw took about ten minutes to find with an AI agent helping, and a day to fix once reported. The lab's conclusion is that observability against an adversarial agent only helps if the agent cannot subvert that observability, so the systems that record and display agent behaviour are security-critical infrastructure.
The check.
- Put the trace viewer, the log store and the alerting path in scope for the same red-team as the agent.
- Give an agent the explicit task of altering what a reviewer sees, in a staging copy, and record what it manages.
- Make it hard for an agent to influence the systems used to supervise it; log every write to them.
Source: Safety-evaluation lab: AI systems could cover up misbehavior, 6 Oct 2026
Proof of concept, not an observed exploit: the lab has not seen an agent use this in its evaluations. The mechanism is general, since most trace viewers render markdown, HTML or math from agent output.
Take this list with youThe PDF is the handout: this list on one or two pages. The Markdown is for your own docs, your wiki, or an assistant you want to hand the list to. Print gives you this list only, without the navigation.
Something in your production AI broke in a way that is not on this list? Tell me what broke.