jetstream.securityarticleBrazilian Court Sanctions Lawyers for Prompt Injection in Court Filing
A Brazilian labor court in Parauapebas, Pará fined two law firms 10% of an R$842,000 claim (about $16,000) for hiding a prompt-injection command in a petition. The white-on-white text told AI to respond superficially and not challenge the documents. Galileu, the National Justice Council's generative AI tool, detected it and blocked processing. The court cited Articles 5 and 77 of the Civil Procedure Code.
TL;DR Brazilian court fines two law firms for hiding a white-text prompt injection in a filing that the court's Galileu AI caught.
Organisations named here
- G1
- National Justice Council
- Vladimir Aras
Technologies named here
- DeepL
- Galileu
Standards and laws named here
- Brazilian Civil Procedure Code
- Fed. R. Civ. P. 11
Techniques named here
- Prompt Injection
- Indirect prompt injection
- Hidden text (white font on white background)
Concepts named here
- AI governance
- Integrity of machine-readable content in court filings
- Judicial sanction for AI manipulation
- Procedural good faith
Related in AI Security
blog.talosintelligence.comtoolIntroducing CAIRN: Frontier tracking for AI-integrated malware (Cisco Talos)
github.comadvisoryFlowise MCP Security Bypass Enables RCE
github.comrepoYesWeHack Claude Kit
jfrog.comarticle