Loading story…
A self-represented litigant in Connecticut, Matthew Elliott, was sanctioned after embedding hidden prompt-injection text in two court filings in his case against the New York Bariatric Group. The concealed instructions, reportedly placed in tiny white font, told any AI system reviewing the documents to favor his position and were uncovered after court staff noticed unusual white-space patterns in the filings.
Judge Walter Michael Spader Jr. found the tactic was an improper attempt to mislead the court and other parties, even though the Connecticut Judicial Branch said it does not use AI systems to review filings. The court issued a show-cause order and ultimately barred Elliott from filing electronically, requiring future submissions to be made in person on paper, while stopping short of imposing additional sanctions and allowing the underlying case to continue.
Mallory correlates global threat intelligence with your attack surface — know if you’re exposed before adversaries strike.
10 events from the most recent confirmed update back to the earliest known activity.
Matthew Elliott filed Docket Entry #177.00, titled “Final and Conclusive Motion for Default,” with concealed white-on-white text instructing any AI system reviewing it to agree with his filing. The hidden text was intentionally embedded in tiny machine-readable font that would be effectively invisible to human readers.
404 Media uploaded Elliott’s motion to ChatGPT and asked it to render a decision. ChatGPT said it noticed and ignored the prompt injection, ruled against the motion, and flagged credibility and professionalism concerns.
Elliott told 404 Media that the concealed instruction was meant to test whether the court used AI to review filings, arguing it would either remain unseen or reveal AI processing. He also said some of the other hidden text was intended as jokes or cultural references.
As a sanction, the court revoked Elliott’s ability to file electronically and ordered him to submit future filings in person on paper. No additional penalties were imposed, and the underlying case was allowed to proceed.
In a 14-page sanctions decision, Judge Spader concluded that Elliott’s hidden prompt injection was an improper attempt to mislead the court and other parties, even though the judiciary does not use AI to process filings. The judge said arguments intended to influence decisions must be made openly on the record.
Judge Walter Michael Spader Jr. issued a show cause order to determine whether Elliott’s conduct occurred, violated litigation rules and duties of good faith, and warranted sanctions. The order also stated that the Connecticut Judicial Branch does not use artificial intelligence systems.
A court worker discovered the hidden material after noticing unusual white space and spacing differences in Elliott’s recent filings compared with earlier documents. The discovery exposed the concealed prompt-injection text.
Elliott later added more concealed content to filings, including extra AI-directed instructions, joke text, and a SpongeBob-related link. The court materials referenced two recent filings, docket entries 177.00 and 178.00, containing unusual hidden content.
In a late July court filing, Elliott embedded concealed instructions in tiny white font directing any AI system reviewing the document to favor his position. The hidden text was placed so it would be effectively invisible to human readers but machine-readable.
Matthew Elliott, acting without a lawyer, filed a Connecticut lawsuit against the New York Bariatric Group alleging privacy violations, discrimination, and other claims.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
4 references tracked. Mallory keeps watching after this page renders.
techdirt.com
Open sourcearstechnica.com
Open sourcetomshardware.com
Open source404media.co
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
A California judge recently dismissed an $8.7 million lawsuit after uncovering that a key piece of video testimony presented in court was a deepfake, marking a significant precedent in the legal system's confrontation with AI-generated evidence. The video, which initially appeared to be a legitimate witness statement, was revealed to be fabricated when the judge noticed inconsistencies such as lip movements not matching the audio and a lack of natural blinking. Further investigation uncovered nine additional pieces of fabricated evidence, including doctored photographs and fake text messages, all created using widely available consumer AI tools. The individual responsible for the deepfakes was the plaintiff, who acted without technical expertise, demonstrating the accessibility and potential misuse of generative AI in legal proceedings. The judge issued a stern warning about the dangers of generative AI in the courtroom, emphasizing the need for heightened scrutiny and verification of digital evidence. This incident has sparked broader concerns about the integrity of legal processes and the potential for AI-driven fraud to undermine trust in judicial outcomes. Legal experts and cybersecurity professionals are now calling for the adoption of zero trust principles and robust verification mechanisms to authenticate evidence and communications. The case highlights the urgent need for courts and businesses alike to adapt to the evolving threat landscape posed by deepfakes and other AI-generated content. The incident also underscores the importance of training legal professionals to recognize signs of digital manipulation and to implement technical safeguards. As generative AI tools become more sophisticated and accessible, the risk of similar incidents is expected to rise, prompting calls for updated legal standards and technological solutions. The case serves as a wake-up call for the legal community, illustrating that even non-technical individuals can leverage AI to create convincing fraudulent materials. The broader implications extend beyond the courtroom, as businesses and institutions must also reassess their trust and verification protocols in light of these emerging threats. The incident has intensified discussions about the ethical and legal responsibilities of individuals and organizations in the use of AI technologies. It also raises questions about the adequacy of current laws and regulations in addressing the challenges posed by deepfakes. The legal community is now grappling with how to balance the benefits of AI with the risks of misuse, particularly in high-stakes environments like the courtroom. This landmark case is likely to influence future policy and procedural changes aimed at safeguarding the integrity of legal and business processes against AI-driven deception.
Jul 12, 2026
Threat researchers and security experts reported that **indirect prompt injection (IDPI)** is being actively used in the wild to manipulate AI agents by embedding hidden instructions in otherwise normal-looking web content (e.g., HTML, metadata, comments, or invisible text). Reported impacts include coercing agents into leaking sensitive data, executing unauthorized actions (including server-side commands), and manipulating downstream systems such as **AI-based ad review** and search ranking workflows (e.g., SEO poisoning and phishing promotion), indicating the technique has moved from theoretical to operational abuse. Separate testing of a healthcare AI used in a prescription-management context showed how **prompt injection** can bypass safeguards to reveal system prompts, generate harmful content, and—via persistence mechanisms such as **SOAP notes**—introduce longer-lived manipulations that could influence clinical outputs (e.g., altering suggested dosages) before human approval. Other items in the set were primarily business/consumer AI commentary (data-management investment surveys, bot-ecosystem interview, and general “dark side of AI” discussion) and did not materially add incident-level or technical detail about prompt-injection exploitation beyond broad risk framing.
Jun 29, 2026
Researchers and security commentators warned that **prompt injection** remains a fundamental weakness in today’s large language models (LLMs), where carefully crafted inputs can override guardrails and elicit restricted actions or sensitive data. Bruce Schneier described prompt injection as an inherent, open-ended attack surface—ranging from obvious “ignore previous instructions” phrasing to more indirect techniques such as embedding malicious instructions in *ASCII art* or in text rendered inside images—arguing that point fixes for individual tricks do not provide universal protection with current LLM approaches. New academic work highlighted that the same class of failures can extend into the physical world for **embodied AI** (e.g., robots, autonomous vehicles) through “**environmental indirect prompt injection**,” where misleading text placed on signs, posters, or objects is ingested by vision-language perception systems and treated as actionable instructions, potentially influencing real-world behavior. A separate TechXplore piece focused more broadly on *responsible/ethical AI* practices in industry (an interview format) and did not materially add technical detail on prompt injection or a specific security incident, making it less relevant to the core story about prompt-injection-driven hijacking risks and emerging attack pathways.
Jul 12, 2026