GPT-5 Debuts: Legal AI Firms Detail Impact & Capabilities
The highly anticipated GPT-5 large language model has officially launched, with several pioneering legal technology firms—Harvey, Spellbook, Legora, and Thomson Reuters—gaining early access and already integrating its advanced capabilities into their platforms. These companies report significant enhancements that promise to reshape how artificial intelligence assists legal professionals.
Winston Weinberg, CEO of Harvey, highlighted GPT-5’s arrival as perfectly timed. He noted that the model’s improved reasoning and ability to orchestrate complex tools will enable the development of more sophisticated workflows. These new workflows can branch, loop, and self-correct, all while ensuring users retain full control through interactive verification. Furthermore, GPT-5’s enhanced long-form writing capabilities are set to transform document generation. Instead of merely producing paragraphs, the model can now compose complete, multi-section documents that maintain structural integrity and internal consistency, adhering to specific legal precedents and drafting conventions. Weinberg credited early access to OpenAI for helping Harvey translate these improvements into tangible benefits for their clients.
Thomson Reuters’ Chief Technology Officer, Joel Hron, emphasized GPT-5’s potential for nuanced legal analysis. He explained that discerning mischaracterizations in legal documents goes beyond simple fact-matching, requiring an evaluation of statements within the full context of a legal brief, cited cases, and their interrelationships—a multi-layered reasoning task spanning multiple documents. Hron noted that previous iterations of OpenAI’s reasoning models had already led to significantly more nuanced and accurate explanations. With GPT-5, Thomson Reuters is building on that foundation, pushing for even deeper contextual understanding crucial for high-stakes legal analysis.
Spellbook, another leader in legal generative AI, has also been collaborating closely with OpenAI on GPT-5. Scott Stevenson, Spellbook’s CEO, indicated that GPT-5 introduces a new level of sophistication in identifying subtle contractual risks. Internal tests by Spellbook revealed GPT-5 as the first model to consistently flag many nuanced, jurisdiction-specific compliance issues that had eluded prior models like Claude 4 and GPT-4o. The model demonstrates a deeper capacity to cross-reference jurisdictional requirements, statutory provisions, and clause language, enabling an unprecedented level of automated legal diligence. Spellbook also observed significant advancements in GPT-5’s fluency with security, compliance, and data protection language, making it particularly effective in refining and expanding enterprise-grade Data Processing Agreements (DPAs).
Beyond identifying issues, GPT-5 is setting a new benchmark for precision editing within complex Word documents—a critical function for transactional lawyers who rarely draft from scratch. Unlike traditional generative models that often regenerate entire sections or struggle with real-world Word formatting, GPT-5 can make surgical edits that preserve surrounding context, reliably populate key fields and remove placeholders, and accurately navigate and revise tabular structures within agreements. Stevenson pointed out that transactional lawyers frequently work with legacy precedents that are often fifty or more pages long, replete with defined terms, interlinked clauses, and embedded tables. He affirmed that GPT-5 is the first model capable of reliably handling these complexities. Early results suggest GPT-5 will power a new class of revision-based workflows where AI functions more like an associate, carefully refining documents rather than merely generating isolated language. GPT-5 is already live in Spellbook Associate, the company’s AI-driven drafting product, with phased deployment underway across Spellbook’s broader review functionality. Marc Manara, Head of Startups at OpenAI, underscored the value of this collaboration, stating that Spellbook’s feedback and early testing would help identify the API’s optimal impact where it matters most for transactional lawyers.
Max Junestrand, CEO at Legora, emphasized his company’s commitment to keeping pace with the exponential improvements in AI capabilities, speed, and intelligence to ensure clients benefit immediately. He highlighted Legora’s early technical decision to be model-agnostic, which means they work closely with all top AI labs, always using the models best suited for the exacting standards of legal work, enabling them to react quickly to the constant changes in the AI landscape.
This launch represents a significant development for the legal and legal technology markets. The capabilities of generative AI tools currently available are fundamentally linked to the underlying large language models. OpenAI’s models, in particular, have been pivotal to the growth of much of the legal AI market since late 2022. As these foundational models improve, especially in text analysis, their advancements directly translate into greater subtlety and accuracy in legal tech tools. This is a clear win for the entire legal sector, which relies heavily on text and its precise interpretation. While Harvey, Spellbook, Legora, Thomson Reuters, and other companies closely involved with the rollout will enjoy an initial head start, it is anticipated that other firms in the sector will soon integrate GPT-5. The relentless push to enhance generative AI, fueled by billions of dollars in investment, ultimately promises better results for the legal market—a segment of the economy that inherently lives and breathes text and its meaning.