Source: Fortune

Anthropic announced on May 12, 2026 that it released more than 20 integrations and 12 role-specific AI plugins for legal workflows, plus a Microsoft 365 cross-app integration embedding Claude across Word, Outlook, Excel and PowerPoint. The rollout is powered by Claude Opus 4.7, which the article reports scored 90.9% on Harvey’s BigLaw Bench, a legal-industry benchmark designed to test LLMs on rigorous legal tasks. Anthropic named Freshfields, Quinn Emanuel, Holland & Knight and Crosby Legal as live users and said products including Harvey, Legora, Solve Intelligence and Eve are built on Claude’s underlying models; more than 20,000 legal professionals signed up for the company’s recent legal webinar.

The article notes recurring problems with AI hallucinations in filings, saying judges have issued sanctions and bar associations warnings, and reporting that rival lawyers found a hallucination in a recent Sullivan & Cromwell bankruptcy filing. Anthropic’s technical response is “grounding,” a connector architecture intended to restrict Claude to live, verified sources such as Westlaw, CourtListener and iManage rather than relying on model memory. Executives quoted — including Gerrit Beckhaus of Freshfields, Christopher D. Kercher of Quinn Emanuel, Jay Madheswaran of Eve, Jake Lauritzen of Legora, and Sonja Ebron of Courtroom5 — praise improved citation faithfulness, long-document consistency, and potential access-to-justice benefits via free connectors to legal aid groups; the piece concludes that whether these guardrails suffice will ultimately be decided in court.