Mistral and Harvey AI: Targeting the legal sector
Using AI in the legal sector is considered sensitive. This is also why Mistral's models are to be used by the legal AI manufacturer Harvey AI in the future.
(Image: Mehaniq / Shutterstock.com)
The models from the French AI company Mistral will soon be usable via the software platform of the US startup Harvey AI. The legal AI platform announced this in a press release. Initially, the Mistral models will be made available to customers in the European Union as part of an early access program. A broader rollout is planned for the coming months.
The San Francisco-based company Harvey AI was founded in 2022. Its main product, the software Harvey, offers LLMs specifically tailored for the legal industry. Through the partnership, Mistral gains access to over 1500 customers in 60 countries who have already integrated Harvey into their business processes.
AI for sensitive matters
The partnership between the two companies is not new. Harvey AI already announced its collaboration with Mistral in a press release in May 2024. At the time, the company emphasized Mistral's commitment to transparency, efficiency, and adaptability. The use of generative AI in highly regulated industries requires the highest level of security and transparency, the company wrote on its website.
As a classically text-heavy domain, the legal system offers great potential for the use of AI. LLMs can help law firms and judicial institutions, for example, to search through hundreds of pages of litigation documents.
A lucrative industry
Law firms also use AI tools for economic reasons: The automation of tasks using AI gives lawyers additional time. This allows them to handle more cases.
Videos by heise
The legal sector is therefore quite competitive among major AI manufacturers. Recently, Anthropic integrated several plugins into its paid AI agent Claude Cowork, which are intended to specialize in legal tasks.
Shadow sides for the judiciary
The increasing use of artificial intelligence in the legal sector also has significant downsides. Often, generative AI is not only used for pure document analysis. Courts worldwide are increasingly struggling with briefs that contain AI hallucinations. The AI tools cite non-existent court rulings, falsify statements, or misrepresent rulings.
According to an online database compiled by legal scholar Damien Charlotin from HEC Paris business school, as of the end of May 2026, there were already nearly 1500 documented court cases in which individuals were sanctioned for submitting documents with AI errors. According to Charlotin, the number of affected court cases has increased significantly since last year. There have already been initial cases in Germany. In early May, a Berlin Regional Court reprimanded lawyers for AI hallucinations.
Some courts in the USA have already introduced labeling requirements for documents processed with AI. However, experts consider these to be impractical and ineffective. Since AI is now integrated into the standard functionality of law firm software, almost every document would be subject to labeling. (rah)
