OpenEvidence and Doximity clash in billion-dollar race to dominate AI for doctors
Summary
The rivalry between OpenEvidence and Doximity has escalated from boardroom competition into duelling lawsuits as both firms race to build the dominant AI assistant for physicians. OpenEvidence — a Sequoia-backed startup valued at about $3.5bn — sued Doximity in June alleging theft of proprietary code and prompt-injection attacks that exposed model system prompts. Doximity, a $13bn public company, has responded with a counterclaim accusing OpenEvidence of false advertising, harassment and poaching employees, and has moved to dismiss the original suit.
Punchy take: this isn’t just bitter startup drama — the outcome could shape legal boundaries around trade secrets, prompt injections and what counts as theft in the AI era. The suits also point to operational risks around clinician identity, patient data exposure and claims about model accuracy.
Key Points
- OpenEvidence sued Doximity claiming the latter impersonated physicians and used prompt-injection techniques to extract trade secrets and training data.
- Doximity counters that OpenEvidence spread misleading claims (including a contested perfect exam score) to harm Doximity’s reputation and recruit its staff.
- The disputes involve wider industry moves: Doximity bought Pathway Medical in August and both companies are aggressively building physician-facing AI tools (scribes, clinical decision support).
- Legal rulings here could set precedents on whether prompt-injection-derived prompts or model instructions are protected trade secrets.
- Both sides also accuse the other of mishandling sensitive information — from alleged misuse of physician identifiers to claims of exposed patient records and contested HIPAA compliance.
Context and Relevance
Why it matters: healthcare AI is a high-stakes market — efficiency gains for clinicians translate into huge commercial value. This case sits at the intersection of technology, privacy and law: courts will likely be pressed to define what constitutes misappropriation when the competitive weapon is a sequence of prompts or model behaviour, not a single file or line of code.
The outcome will influence investor caution, product claims around accuracy and hallucinations, hiring/poaching tactics, and how startups protect model prompts and system instructions. Clinicians and healthcare organisations should watch how data-handling and HIPAA-related allegations are treated, because regulatory exposure could follow any finding of careless data practices.
Why should I read this?
Short version: because this fight is where the rules for AI in medicine are getting written — and the stakes are huge. If you care about AI in healthcare, legal risk, or which tools doctors will actually trust and use, you want to know who’s winning the argument and why. Plus, it’s full of juicy specifics (prompt-injection claims, alleged recruitment messages, contested accuracy bragging) that tell you how cutthroat this market has become.