OpenAI Pushes Codex Beyond Code Into Finance and Legal, Squaring Off Against Anthropic's Claude for Legal
OpenAI is extending Codex into finance and legal work, weeks after Anthropic expanded Claude for Legal with 12 plugins and 20-plus integrations.
Overview
The race to embed generative AI inside specific professional industries has opened a new front. OpenAI is expanding Codex, originally a software-coding tool, into business verticals such as finance and law, according to The Tech Portal. The move follows Anthropic, which weeks earlier mounted what it called its most expansive push yet into legal workflows, as reported by Fortune. Two of the largest model developers are now competing directly for the same high-value professional customers — and doing it through industry-specific products rather than general-purpose chatbots.
What We Know
OpenAI is building out a set of role-specific Codex plugins aimed at knowledge workers. The company confirmed that additional plugins are already under development, including Corporate Finance, Private Equity Investing, Marketing Strategy, Strategy Consulting and Legal, according to The Tech Portal. Finance is the furthest along: the public-equity product is designed to help investors analyze earnings, compare companies, track signals and evaluate whether an investment thesis is strengthening or deteriorating, the same outlet reported. To power that work, The Tech Portal reported that OpenAI integrated access to institutional-grade information providers including Moody’s, FactSet, LSEG, PitchBook, Daloopa, Datasite and S&P. The expansion intensifies OpenAI’s competition with Anthropic, which has been aggressively targeting the same professional markets, the outlet noted.
Anthropic got there first in the legal vertical. On Tuesday, May 12, the company announced its most expansive push yet into legal workflows, releasing more than 20 new integrations with the tools law firms already rely on, according to Fortune. That release included 12 role-specific AI plugins covering everything from M&A due diligence to employment handbook drafting, and a cross-app integration with Microsoft 365, Fortune reported. TechCrunch reported that the same day Anthropic launched a host of new chatbot features designed to provide automated assistance to law firms, building on Claude for Legal, the law-focused plug-in that launched earlier this year. The new tools are designed to help law firms automate specific clerical functions — things like document search and review, case law resources, deposition prep, document drafting, and other related areas, TechCrunch reported.
The products are already in use at large firms. Fortune reported that Freshfields, Quinn Emanuel Urquhart & Sullivan, Holland & Knight, and Crosby Legal are among the firms applying Claude to live matters.
The Money Behind Legal AI
The vertical is drawing heavy investment beyond the two large model developers. In March, the AI law startup Harvey, which uses agentic AI to automate legal workflows, raised $200 million at a valuation of $11 billion, according to TechCrunch. A rival startup, Legora, raised a $600 million Series D, the same outlet reported. An Anthropic spokesperson framed the urgency this way to TechCrunch: “The legal sector is facing mounting pressure to adopt AI, and the firms and in-house teams that move are pulling ahead fast.”
The Reliability Problem
The push into law arrives against a backdrop of well-documented reliability failures. Judges have been issuing sanctions, bar associations have been issuing warnings, and lawyers across the country have been caught submitting briefs containing citations to cases that never existed, according to Fortune. Fortune reported that Sullivan & Cromwell, a white-shoe law firm, was found by rival lawyers to have included a hallucination in a bankruptcy court filing.
Vendors are pointing to benchmark scores as evidence of progress. Fortune reported that Claude Opus 4.7 scored 90.9% on Harvey’s BigLaw Bench, which it described as the legal industry’s most closely watched AI benchmark. But practitioners caution that high scores do not eliminate the core risk. Jay Madheswaran, CEO and co-founder of Eve, told Fortune: “In litigation, an authoritative-sounding hallucination is worse than no answer.”
Analysis
The contest now taking shape is less about which model scores highest and more about which company can wire its model into the software lawyers and bankers already use every day. Anthropic’s May release leaned heavily on integrations — more than 20 of them, plus a Microsoft 365 tie-in, per Fortune — while OpenAI’s finance plugins lead with connections to institutional data providers, per The Tech Portal. In both cases, the differentiator is the ecosystem of connectors and trusted data, not the raw model.
That framing also raises the stakes on reliability. A general chatbot that occasionally invents a fact is an annoyance; a legal assistant that invents a case citation can draw a sanction, as the filings documented by Fortune show. As OpenAI moves Legal from its roadmap into a shipping product — it remains under development, per The Tech Portal — the question for buyers will be whether vertical-specific tooling and curated data sources can keep authoritative-sounding errors out of work that carries professional and legal consequences.
What We Don’t Know
OpenAI has not announced pricing, a launch date, or a final product name for its Legal plugin, which The Tech Portal describes only as under development. It is also unclear how the two companies’ legal offerings will compare on accuracy in real-world use, as published benchmark figures such as the BigLaw Bench score reported by Fortune measure performance under controlled conditions rather than in live litigation.