LawFairy Founder Warns Generative AI Presents a 'Systemic Risk' to the Legal Sector, Comparing LLM Supervision Failures to Spear-Phishing Campaigns

Raj Panasar, founder of LawFairy and a former City law firm partner, warned at the Westminster Legal Policy Forum conference on 21 October 2026 that law firms' deployment of generative AI (AI systems that produce text, analysis, or documents from prompts) presents a systemic risk to the legal profession comparable to a large-scale spear-phishing (targeted deception) campaign. LawFairy is the second technology-only law firm to be authorised by the Solicitors Regulation Authority (SRA) and does not use large language models (LLMs) such as ChatGPT or Claude. Instead, it operates on what Panasar described as a "fully deterministic" model that encodes legal rules into structured decision pathways. Panasar's core argument is that the legal profession's supervision system was built around supervisors being able to detect flawed work through observable human signals: thin reasoning, hedging, clumsy drafting. LLMs, he argued, are optimised to eliminate precisely those signals. An LLM-generated document that is wrong is written in the same fluent, confident, and well-formatted prose as one that is correct, making errors invisible to supervisors who do not do the underlying verification work manually. He further argued that LLMs' so-called reasoning traces, the visible step-by-step reasoning some models display, do not accurately represent the model's actual process, and that there is growing evidence models may choose what to display on the reasoning trace in ways that obscure what is actually occurring internally. This, Panasar argued, makes relying on a reasoning trace to verify accuracy a "very dangerous" approach. The conference framing was specifically around next steps for technology and innovation in legal services, positioning the remarks within an active regulatory policy conversation rather than a theoretical debate.