Browse / Indirect Prompt Injection / Real-world incidents
Real-world incidents
4 records in Indirect Prompt Injection.
- Hidden Prompt Injection in Brazilian Labor-Court Petition Reportedly Tried to Manipulate Galileu Hidden instructions embedded in a labor-court petition told Brazil's Galileu AI tool to contest the case superficially and ignore documents; the tool detected and blocked them.
- Unknown Actor Reportedly Exploited Cline's Claude-Powered GitHub Issue-Triage Workflow and Published an Unauthorized Cline CLI Package An attacker reportedly injected instructions into Cline's Claude-powered GitHub issue triage, combined with cache poisoning, to steal an npm token and publish an unauthorized package.
- Preprints Reportedly from Researchers from Multiple Universities Allegedly Contain Covert AI Prompts Researchers reportedly hid prompts in arXiv preprints, using white or tiny text, instructing AI tools to give only positive peer reviews to any reviewer relying on AI.
- Amazon Alexa Responding to Environmental Inputs Alexa devices acted on voice commands in TV commercials and broadcasts, such as ordering products, so environmental audio effectively injected instructions into the assistant.