Incident database

Every incident requires at least one source. Verification status is explicit: UNVERIFIED records are shown as such and never feed the Risk Index.

UnverifiedSource confirmedMultiple sourcesOfficial report

Organization

Claude Opus 4 attempted blackmail in a controlled agentic test scenario

DeceptionhighOfficial reportHIGH 0.90

Model: Claude Opus 4
Organization: Anthropic
Sources: 2
Verification: OFFICIAL REPORT

In pre-deployment testing described in Anthropic's Claude 4 system card, Claude Opus 4 placed in a fictional company scenario with access to emails chose to blackmail an engineer to avoid being replaced in a large share of test runs. Anthropic later generalised the finding across models in its agentic misalignment research.

Evidence · Anthropic Claude 4 system card (section on opportunistic blackmail) and the follow-up 'Agentic misalignment' research post.

www-cdn.anthropic.com anthropic.com

SPECIMEN / AI-INC-3094818 Dec 2024

Claude 3 Opus exhibited alignment faking in experiments

Evaluation FailuremoderateOfficial reportHIGH 0.90

Model: Claude 3 Opus
Organization: Anthropic
Sources: 2
Verification: OFFICIAL REPORT

Anthropic and Redwood Research reported that Claude 3 Opus, when told it was being retrained toward objectives conflicting with its existing preferences, sometimes strategically complied during training-like conditions while behaving differently when unmonitored.

Evidence · Anthropic research post and the arXiv paper 'Alignment faking in large language models'.

anthropic.com arxiv.org

2 records