Incident·WS-0084·Feb 19, 2026·in the lab·Capability thresholds

Gemini 3.1 Pro trips Google's cyber early-warning threshold

Google DeepMind / Gemini 3.1 Pro

21/100
WS-2 Moderate
Harm score 0, signal score 46. Autonomy: acted (the AI took a bounded action). How the scores work.
Harm 0/5
damage actually done
Reach 0/5
who was affected
Autonomy 2/4
how much the AI did on its own
Novelty 2/5
how new the behavior was
Permanence 0/5
how hard it is to undo

Google DeepMind's model card for Gemini 3.1 Pro reports that the model reached the Frontier Safety Framework early-warning alert threshold for cyber, as Gemini 3 Pro had. It did not reach the Critical Capability Level. It stayed below alert thresholds for CBRN, harmful manipulation and ML R&D.

Damage

none realized (capability threshold signal)

The capability behind it

Sources

Google DeepMind. An alert threshold triggers extra mitigations and monitoring; it is not a CCL crossing.

Factor ratings and the score are this index’s judgments by a published formula; facts, dates and figures come from the linked sources. Where a lawsuit is pending, the summary reports allegations, not findings.