Georgia ADA Used AI to Draft Court Brief, Cited at Least 10 Fake or Misquoted Cases

·5 views

On March 18, 2026, Georgia Supreme Court Chief Justice Nels Peterson publicly confronted Clayton County Assistant District Attorney Deborah Leslie over at least ten fabricated or misrepresented case citations in a brief opposing a new trial for convicted murderer Hannah Payne. Leslie had used AI to draft the filing and, by her own admission, failed to verify the generated citations before submission. Clayton County DA Tasha Mosley confirmed the AI use in a letter to the Chief Justice on March 27, and Leslie was subsequently suspended and referred to the State Bar of Georgia. The incident is part of a broader pattern tracked by legal scholar Damien Charlotin, whose AI Hallucination Cases database had logged over 1,200 such incidents worldwide by spring 2026, with roughly 800 from the United States alone. Experts warn that AI-generated legal hallucinations are becoming a structural problem in modern legal practice, with the documented rate continuing to rise.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Sony CEO Hiroki Totoki Drives Major Overhaul of the Japanese Tech Giant

Hiroki Totoki, a long-time Sony veteran, has taken the helm as CEO and is leading a significant restructuring of the company. Totoki spent decades rising through Sony's ranks before assuming the top leadership role. His reform efforts aim to modernize and streamline the conglomerate, which spans entertainment, electronics, and financial services. The overhaul reflects broader pressures on Sony to improve profitability and adapt to a rapidly changing global market.

0 comments Read more at Hacker News

ProgrammingHacker News ·

Developer builds personal root DNS to escape domain name rental fees

A developer has built their own root DNS system as an alternative to the traditional domain name infrastructure. The project was motivated by frustration with the recurring cost of renting domain names through conventional registrars. By running a personal root DNS, the developer gains full control over name resolution without ongoing fees. The project was shared on Hacker News, where it attracted community attention and discussion. It highlights a growing interest among technical users in self-sovereign internet infrastructure.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Self-taught solo dev builds doc-code drift detection system to catch governance failures

A self-taught developer and coffee roaster in Taiwan built a custom governance system after spending eight months coding an ERP platform for his coffee business using AI tools. In April 2026, during his most intensive development month with 348 git commits, he repeatedly encountered the same bugs resurfacing across multiple code locations despite repeated fixes. He traced the root cause to a disconnect between his written specifications and actual code behavior — what he calls 'drift' between documentation and implementation. To address this, he created three governance artifacts: a CI script to block rule-violating pushes, a tech-debt backlog file, and reusable 'skill files' to preserve rules across sessions. The workflow enables prose documentation and code to mutually audit each other, offering a structured approach for solo developers to manage governance without a dedicated team.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Encrypted AI Reasoning Traces Found Leaking Secrets Across Users and Models

Security researchers disclosed in August 2026 that encrypted reasoning traces used by OpenAI, Anthropic, and Google could be replayed outside their original session context, including across different users and weaker AI models. The flaw allowed hundreds of real secrets — such as API keys, passwords, and access tokens — to be extracted from internal chain-of-thought blocks that never appeared in visible model outputs. Agentic AI workflows were particularly exposed, as sensitive data like credentials often enters a model's reasoning process without ever surfacing in the final response. Standard security tooling failed to detect the leaks because it is built to inspect visible input-output streams, not encrypted internal reasoning blocks. Researchers noted the core issue was that encrypted reasoning objects were not cryptographically bound to a specific session, user, or model identity, making encryption effectively a form of obfuscation rather than true access control.

0 comments Read more at DEV Community