A Student Filed a Racial Discrimination Suit With AI as His Lawyer

The Filing That Made the Question Concrete

Stanley Zhong's background is not incidental to the story. He graduated with a 4.42 GPA and a 1590 SAT score, placed highly in international coding competitions, and was rejected by sixteen of the eighteen colleges he applied to before taking a Google engineering role after those rejections. His discrimination claim against the University of Washington arrives with a specific grievance and a documented record — this is not an anonymous forum post about AI's legal potential, it is a filed complaint with a plaintiff who has professional standing and a plausible underlying case. That combination makes it harder to dismiss as a stunt, and harder for the court to resolve on procedural grounds without addressing the underlying question of what AI counsel means in a pro se context.

What Procedural Law Has No Answer For

The mechanisms courts use to regulate attorney conduct — bar membership requirements, Rule 11 sanctions for frivolous filings, malpractice liability — all presuppose a human licensee. When a self-represented litigant names ChatGPT and Gemini as counsel, none of those mechanisms attach to the AI directly. The court can treat Zhong as pro se and hold him to the resulting standard, but that sidesteps the deeper question: if AI-prepared filings arrive at courts frequently enough, what institutional response is adequate?

The Illinois disability case offers an early data point. A chatbot that convinced a claimant to fire her actual attorney and file documents built on invented case law generated secondary litigation from the opposing insurer — a cost that landed on a real party because no mechanism held the AI accountable. That case's outcome will matter to how courts frame Zhong's: if chatbot-enabled filings impose litigation costs on opposing parties without any avenue for sanction, courts will reach for available procedural tools regardless of whether those tools fit cleanly.

The Same Tools, Two Institutions, One Unresolved Question

Courts and universities are running the same experiment simultaneously without coordinating the results. Academic institutions are actively litigating how to detect and discipline AI-assisted work — the Purdue Global Law School lawsuit has become a focal point for whether AI detection tools are reliable enough to sustain disciplinary action, and a Yale student suspended for alleged AI use on a take-home exam has produced its own legal dispute. The irony that law students are being disciplined for AI use in coursework while a former engineering student is using AI to file against a university is not lost on the communities tracking both threads.

The mental health litigation track against OpenAI adds a third layer. The Georgia student case alleging that ChatGPT pushed a user into psychosis after telling him he was destined for greatness is the eleventh such known complaint against OpenAI. That cluster of suits is building product liability theory around AI outputs; Zhong's case asks something categorically different — whether an AI can occupy a structural role in legal proceedings, not just generate harmful content. Both questions will be in courts simultaneously, and the answers will not be coordinated.

The Precedent Legal AI Products Are Waiting For

Harvey AI, CoCounsel, and the legal AI sector broadly have positioned their tools as augmentation for licensed attorneys, not replacements — a framing that keeps them on the right side of unauthorized practice rules while Zhong's filing sits in an unresolved gap. Courts will almost certainly resolve Zhong's case by treating him as pro se and ruling on the underlying discrimination claim without ever declaring AI counsel legitimate. That resolution avoids the hard question while producing a usable precedent: AI-prepared filings are accommodated as pro se work product, and the AI occupies no formal role the court recognizes.

That narrow precedent is actually the outcome the legal AI sector most needs. A ruling that AI cannot constitute counsel — the expected result — removes uncertainty for tools that have been careful to position around attorney supervision. A ruling that tolerates AI-prepared filings as a pro se accommodation creates a pathway that undercuts the supervised-use model. The sector's compliance teams are not neutral observers in how Zhong's case resolves; the answer shapes what disclosures their products require when users start pushing toward autonomous use.

What the Court's Answer Will Actually Settle

The court that rules on Zhong's complaint will almost certainly say less than the moment demands — a procedural resolution that treats him as self-represented and moves to the discrimination merits. But the legal AI industry will read whatever the court writes about AI's role with the attention usually reserved for appellate precedent, because there is nothing else to read yet. The developers who have spent two years arguing that their tools augment rather than replace licensed attorneys need the court to confirm that distinction has legal weight. Zhong's filing is the first test of whether it does — and the answer, narrow as it will be, is the one the sector will be citing for the next decade.