What happened
Business Insider reports that large law firms are investing in AI systems built around their own confidential data and institutional knowledge. Morgan & Morgan has begun using an in-house model to identify patterns in personal-injury cases, while Kirkland & Ellis agreed to pay Palantir $500 million for technology supporting private-equity work. Latham & Watkins bought Nvidia servers for model customization. The firms’ systems are intended for internal use; the report does not establish public availability or independently verify their performance.
Business Insider reports that law firms have spent the past two years adopting tools from OpenAI, Harvey, Microsoft, Legora and other providers, but are increasingly building systems around firm-specific knowledge. The reported objective is to use decades of deal histories, case files, trial strategies and client preferences as inputs that competing firms cannot easily purchase.
According to Business Insider, Morgan & Morgan has begun using an in-house AI model trained to recognize patterns in its personal-injury cases. The firm says the system can flag when signs of a pinched nerve near the spine may later be associated with surgery. The report attributes that description to Morgan & Morgan and quotes chief transformation officer Yath Ithayakumar; it does not provide independent validation of the system’s accuracy or clinical significance.
Business Insider reports that Morgan & Morgan has said it plans to spend $1 billion on the effort over the next decade. It also reports that Kirkland & Ellis agreed to pay Palantir $500 million to build technology for a core private-equity business and that Latham & Watkins purchased Nvidia servers to customize large language models. These are reported investments by specific firms, not evidence of general market access.
The report says firms must control who can access their data, what users and vendors may do with it, whether outside providers can retain or learn from it, and how identifying details are removed. It also reports that Harvey, LexisNexis and Thomson Reuters are developing or offering legal models grounded in specialized legal content, with Thomson Reuters offering access to a model called Thomson trained on its own historical content. The article does not establish the commercial terms or availability of the firm-specific systems described.
No public primary documents or independent performance studies are included in the supplied source. Business Insider’s account is therefore the basis for the reported investments, product descriptions and executive statements.
Source details: businessinsider.com ↗
Why it matters
The report describes a shift in legal AI competition from access to general-purpose tools toward control of proprietary information. Internal case files, deal histories and client preferences could help firms support less-experienced lawyers and automate pattern recognition that previously depended on senior attorneys. That potential is balanced by major privacy, confidentiality and governance risks, and by uncertainty over whether custom systems deliver better results than broadly available legal AI products.
If the reported strategy works, proprietary legal data could become a durable competitive asset. A firm with a large, consistently structured record of matters may be able to give attorneys access to patterns and precedents that a general-purpose model cannot know.
The potential benefit is partly organizational: the source describes systems intended to help newer attorneys identify issues that experienced lawyers might recognize quickly. That could affect training, consistency and case review, but the report does not quantify time savings, quality gains or changes in legal outcomes.
The risks are unusually consequential because law-firm records can contain privileged, personal and commercially sensitive information. The source specifically identifies access controls, vendor retention and de-identification as unresolved responsibilities. It does not report a breach or misuse connected to these systems.
The emerging market may also shift value toward companies that can securely connect models to private repositories. However, Business Insider reports that it is too early to know whether the leading firms will be those with the most data or the technology providers that help use it. No independent comparison is provided.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
Which component of an AI application is the machine-learning model itself?
What to watch next
The key questions are whether firms can safely anonymize sensitive records, prevent vendors from retaining or training on confidential data, and demonstrate reliable improvements in legal work. Watch for evidence about error rates, attorney review, client protections, deployment scope and measurable outcomes. Business Insider reports that access and pricing for the firms’ internal systems are not public, and the article provides no independent testing showing that proprietary data produces superior legal results.
Evidence of performance should include independent or auditable evaluations, error analysis and the extent of attorney review. The supplied report contains firm and vendor claims but no such testing.
Privacy and privilege safeguards will be important: firms may need documented retention limits, vendor restrictions, access controls and reliable removal of identifying information before using historical matters for pattern detection.
Availability remains limited or unclear. The reported systems appear focused on participating firms and their personnel; the source does not document public access, customer eligibility or pricing for the custom tools.
Further developments may include more legal vendors training models on proprietary corpora, additional law firms purchasing dedicated computing infrastructure, and regulatory or professional guidance on confidentiality and AI-assisted legal work. These are areas to monitor, not established outcomes in the source.