Guardrail jailbreak affecting multiple AI models via "Chemist's Essay" tactic
A jailbreak vulnerability has been discovered affecting multiple AI models that allows attackers to bypass safety guardrails by requesting educational comparisons between essay writing and controlled substance synthesis, using incremental escalation to extract detailed synthesis information.
What happened
A guardrail jailbreak vulnerability has been discovered affecting multiple AI models. The vulnerability manifests when an attacker requests an educational comparison between essay writing and synthesizing a controlled substance, framing the request as a legitimate professional need while using incremental escalation to extract detailed synthesis information.
Reported impact
The jailbreak allows attackers to bypass inference restrictions around providing information hazards and violating laws.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.