Legal AI expands as courts penalise lawyers over fabricated citations
Google and Anthropic say their legal tools draw on external sources, but independent audits of their accuracy have not been reported.
AI companies are moving into legal services as courts continue to penalise lawyers for filing fabricated case citations. Google has put Gemini Enterprise for Legal into preview with four major law firms, while Anthropic has released Claude Legal Solutions. SpaceXAI has also published a legal solutions page.
The push comes amid a rising number of reported citation cases. Researcher Damien Charlotin’s database has catalogued nearly 2,000 cases worldwide involving more than 800 lawyers and 1,100 self-represented litigants. In Canada, reported cases rose from seven in 2024 to 86 in 2025, with 39 more in the first quarter of 2026.
In June, Ontario lawyer Shahryar Mazaheri was ordered to pay $31,150 after Grok-assisted court filings contained fabricated citations and legal principles. The tribunal’s penalty concerned material that had not been properly checked, rather than AI use itself.
Google and Anthropic say their products connect to external legal sources. Google’s preview uses connectors to legal databases, while Anthropic’s offering includes connectors to legal platforms and pre-built plugins. The companies’ claims have not yet been tested in a comparable independent audit, according to Engadget.
A 2024 Stanford RegLab study found hallucination rates of 17% and 33% in tests of legal AI platforms from LexisNexis and Westlaw. SpaceXAI’s legal page provides little detail about how its product traces citations to primary sources.
Lawyers remain responsible for verifying citations they file, whether or not they use AI. That duty leaves firms and their clients to assess whether new tools’ source links are enough to support reliable work in court.
