Jurisdiction
Everlaw and DISCO each published model documentation for their AI review features, listing training data provenance, evaluation methodology and known limitations. Neither disclosed model weights or vendor. The publications follow customer requests for documentation to support internal AI governance reviews. Both companies said the documents will be updated with each model change.
A Northern District of California judge issued a standing order requiring any filing prepared with generative AI to identify the tool, the portions affected, and the human who verified them. The order applies to all civil matters on the docket and takes effect for filings made after 1 August. Certification is required with each filing rather than once per case. Counsel who cannot make the certification must instead file a short statement explaining why, before the applicable filing deadline.
The Ninth Circuit adopted a circuit-wide rule requiring counsel to certify that every authority cited in a brief has been checked against a primary source. The rule does not require disclosure of AI use. Judges may strike a brief and order re-filing at counsel's expense. The court said it chose verification over disclosure because the risk is unverified content, not the tool.
An academic team published a benchmark measuring citation accuracy in AI legal research tools, testing seven commercial products against a set of two thousand questions drawn from state and federal practice. The paper reports error rates varying by an order of magnitude between products and finds that retrieval-augmented systems reduce but do not eliminate fabricated citations. The authors released the question set and scoring code, and invited vendors to submit corrections before a second round of testing later this year.
Read the long piece on LexRegisterTwo US law firms filed breach notifications describing ransomware incidents affecting client matter data, together covering roughly forty thousand individuals. Both notifications state that data was exfiltrated before encryption. Neither firm named the vendor or system involved. State attorneys general published the notifications in their breach registries, which is how the incidents became public.
A Southern District of New York judge imposed sanctions on two lawyers whose brief cited six cases that do not exist, referring both to the state bar and ordering them to notify their client. The court found the citations were generated by an AI research tool and filed without verification. Sanctions include a monetary penalty and mandatory continuing education. The opinion sets out a verification standard that the court says it will apply to future filings across its docket.
Corrected . An earlier version of this card said the penalty was $7,500. Reuters and Law360 report $5,000; Bloomberg Law reports $7,500. The summary now states the sanction without a figure and the key point carries both.
Read the long piece on LexRegisterA proposed class action filed in the Northern District of California alleges that marketing claims about hallucination-free legal research were false and misleading. The complaint seeks refunds for subscription fees paid by small firms and solo practitioners. It pleads claims under California's unfair competition law. No defendant has answered, and the docket shows a scheduled hearing on the motion to dismiss in October.
Read the long piece on LexRegisterAccel led a $550 million Series D in the Swedish legal AI company Legora, tripling its valuation to $5.55 billion. The company said the round funds expansion in the United States and continued hiring for its agentic drafting product. Fourteen outlets reported the round on 10 March; none reported secondary-sale terms.
A District of Columbia judge set out an authentication standard for evidence alleged to be AI-generated, requiring the proponent to show provenance through metadata, chain of custody or testimony from someone with direct knowledge. The order declines to adopt a presumption either way. It applies to a single case but the court published it as guidance for the district.
A federal district judge issued a standing order requiring parties to certify whether generative AI drafted any portion of a filing and to confirm that every citation was checked against an authoritative database. The certificate must accompany each pleading. The order follows comparable requirements adopted judge by judge since 2023 and is enforceable by striking the filing.
NIST released a draft profile applying the AI Risk Management Framework to legal services, mapping the framework's govern, map, measure and manage functions onto tasks such as document review and legal research. The draft is open for comment for sixty days. It is guidance rather than a rule, and carries no independent enforcement mechanism.
Litera and DocuSign released a joint integration linking drafting and contract execution, allowing a document to move from a Litera template to a DocuSign envelope without leaving Word. The integration is available to customers of both products at no additional licence cost. Neither company disclosed how many joint customers the integration is expected to reach.
A Northern District of Illinois judge approved a model discovery protocol governing the use of AI-assisted review, requiring parties to disclose the tool, the validation method and the recall achieved. The protocol does not require disclosure of seed sets or training decisions. It was drafted with input from both sides' e-discovery vendors and is offered for adoption in other matters.
A study of bar examination performance found that candidates permitted to use a generative AI assistant scored higher on multistate performance test items but not on multiple-choice questions. The authors caution that the sample was small and drawn from a single jurisdiction's practice examination. Examiners have not indicated any change to permitted materials.