A Brazilian labor court fined two law firms after finding that lawyers deliberately hid a prompt-injection instruction in an employment-court filing. The text was rendered white on a white background, making it invisible during ordinary reading. It instructed any AI system reviewing the petition to respond superficially and avoid challenging its documents. The judge discovered it while using Galileu, a generative AI tool used by Brazil’s National Justice Council; after the font color was changed, the command became visible. The underlying claim sought R$842,000, and the court imposed a fine equal to 10% of that amount, approximately US$16,000.
The court held that the lawyers’ conduct aimed to bias an AI-generated response or draft judgment against the defendant, undermining both the proceedings and the administration of justice. It grounded the penalty in Brazilian procedural rules requiring good faith, loyalty, and truthfulness, and prohibiting conduct that obstructs judicial work. The decision treats machine-readable content as part of a filing’s integrity, not as a loophole separate from its visible text, and may be the first judicial ruling to address deliberate prompt injection against a court system. The AI reportedly detected and blocked the suspicious content, illustrating both the risks of documents processed by AI and the value of safeguards that monitor for manipulation.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
China’s STAR 50 technology index has fallen 40% since June, wiping trillions from the country’s stock markets as investor anxiety grows over AI. Weak profits and Western threats are raising doubts about the sustainability of Beijing’s push to compete with the US for AI dominance. (theguardian.com)
Software engineer and consultant Eric Elliott says his mentoring and consulting business collapsed as generative AI reshaped demand, leaving him homeless despite decades of experience and AI expertise. He argues that learning AI alone cannot protect workers from displacement and warns of broader economic fallout. (medium.com)
The spread of corporate AI is pushing more tech workers to reconsider their long-standing resistance to unions. They are turning to collective bargaining to challenge how companies deploy AI and protect workers’ interests. (theguardian.com)
![Honest Government Ad – AI Data Centres [video]](https://hntoday.pixelshot.io/screenshots/20261101/a078d3db411845f0_tn.jpeg)
A satirical “honest government” ad focuses on AI data centres. The supplied page provides no details about the video’s specific claims. (youtube.com)
Ed Zitron’s predictions were wrong in 25 of 27 cases reviewed by the author. The proposed “Inverse Zitron” strategy is to invest against his calls. (chalmovsky.com)
More capable AI models can now build and test production features from prompts, even through mobile tools. Software engineers will increasingly focus on defining tasks, reviewing code, and validating results as AI writes more of it. (newsletter.pragmaticengineer.com)
Today's best Hacker News stories, summarized and screenshotted, one email a day.