Mar 1, 2023, 10:00 AM ETInterview

The Robot Lawyer Resistance with Joshua Browder of DoNotPay

Founding Story: Joshua Browder founded Do Not Pay in 2016 after accumulating ~30 parking tickets in London and the US while unable to afford fines, realizing legal advocacy could be automated for common issues.
Scale and Reach: The company has successfully resolved over 2 million cases and currently serves hundreds of thousands of active subscribers.
Mission: To act as a "general counsel for consumers," utilizing technology to help individuals fight corporations and bureaucracy they cannot afford to litigate manually.
Product Scope: Offers 100+ distinct use cases, ranging from parking ticket disputes and subscription cancellations to obscure rights like free raffle entries and pet recovery forms.

Phase 1 (Document Automation): Initially relied on pre-written legal templates and decision trees to generate letters, effectively automating low-complexity tasks.
Phase 2 (AI Integration): Transitioned to true AI to handle dynamic, multi-turn negotiations (e.g., medical bills) where simple templates fail to address complex back-and-forth disputes.
Medical Bill Automation: Leveraging the No Surprises Act (2022), the system scrapes hospital price databases (1.4 million rows of data) to negotiate "good faith" discounts by comparing patient bills against local market rates.
Refund Success Rates: Automated letters regarding in-flight Wi-Fi failures achieve refund success rates of over 80% by citing FTC statutes.
Training Methodology: Models are fine-tuned using specific legal outcomes and historical data rather than relying on generic LLMs, ensuring accuracy in specific jurisdictions and laws.

High-Value Focus: Prioritizing cases where potential savings are 10x the average, specifically targeting medical bills (up to $10,000) rather than low-value parking fines.
Terms & Conditions Analyzer: A new AI tool that analyzes contracts, leases, and terms of service to identify non-standard or predatory clauses for the average consumer.
Spam Caller Litigation: An automated system to sue telemarketers under the Telephone Consumer Protection Act (1991) for $1,500 per call, using "trap" credit cards to extract merchant details for litigation.
Class Action Vision: Aims to enable "instant class actions" where AI monitors corporate actions and automatically files claims, bypassing the decades-long delays typical of current litigation.
Cross-Border Capability: AI is utilized to apply foreign laws (e.g., Mexico's 5-day timeshare cancellation law) to consumers in jurisdictions where human lawyers lack expertise.

Unauthorized Practice of Law (UPL) Pushback: The legal industry has threatened prosecution over an experiment to use AI to whisper advice to drivers in physical courtrooms via bone-conduction glasses.
Bar Association Stance: State bar associations (e.g., in California) view AI representation as violating rules written for human attorneys, despite Browder's argument that the human is representing themselves with AI assistance.
Judicial Resistance: Judges, particularly in traffic courts, often exhibit bias against remote or AI-assisted representation, with some expressing prejudice toward Zoom appearances.
Corporate Countermeasures: Large corporations (e.g., Comcast) now employ their own AI to negotiate with consumers, turning the interaction into an "AI vs. AI" battle.
Legislative Risks: Browder warns of potential "anti-AI laws" that could ban unverified bots, though he argues current consumer rights remain enforceable even if AI authorship is disclosed.

Emotional Limitations: AI is explicitly excluded from cases requiring emotional nuance, such as asylum hearings, divorce proceedings, or criminal defense, where human empathy and presence are legally and practically necessary.
Hallucination Risks: The company actively guards against AI fabrication of facts (e.g., lying about internet outages), which could lead to perjury charges, by training models to adhere strictly to verified facts.
Excessive Verbal Output: Models are being retrained to avoid "talking too much" or responding to rhetorical questions in court, which could annoy judges and compromise cases.
Quality Control: Do Not Pay asserts that errors in consumer rights cases are negligible compared to human lawyer errors, citing that human lawyers can be "drunk" or steal client funds, whereas AI remains objective.
Market Defense: Browder argues that the legal establishment's protectionism is a defense against job displacement, as ~80% of legal work involves document assembly that AI can automate.

Deflationary Pressure: AI automation in customer service and legal administration could reduce corporate overhead by ~10%, potentially lowering prices for all consumers, not just legal fees.
Legislative Writing: Browder predicts AI will soon be used by congressional staff to draft legislation, including laws governing AI itself.
Inevitable Modernization: Citing the 80% of people who cannot afford legal counsel, Browder expects courts to eventually mandate or allow AI assistance to prevent justice system collapse.
Supreme Court Experiment: The company previously offered a $1 million prize to any lawyer willing to use AI in the Supreme Court to demonstrate capability, though no major firm accepted.
Long-term Vision: A future where consumers are unaware of violations because "robot lawyers" instantly resolve refunds and legal disputes in the background without user intervention.