Legal News · Region
Global legal news
Stories without a single home market.
- +3Legal AI Models
Google Cloud launches Gemini Enterprise for Legal in preview
Google Cloud has put a legal product of its own on the board. Gemini Enterprise for Legal is a legal layer inside the wider Gemini Enterprise app rather than a separate application, pairing task-specific skills for contract review, regulatory horizon scanning, DSAR fulfilment and motion redaction with MCP connectors into iManage, NetDocuments, Docusign, Everlaw, RelativityOne, Harvey, Legora, Thomson Reuters HighQ and CourtListener. It is preview only, with Cleary Gottlieb, Freshfields, Weil and Williams & Connolly named as launch customers, so buyers now weigh a hyperscaler that sits partly on top of the legal specialists they already licence.
SourceGoogle Cloud BlogGoogle Cloud Press Corner (PR Newswire)FactorArtificial LawyerReutersThe American Lawyer
- Model Releases
Thomson Reuters launches Thomson, its own legal LLM
Thomson Reuters now runs a model it owns, named Thomson, instead of relying only on outside frontier vendors. It starts from an open-source base and is specialised on the company's own libraries, Westlaw and Practical Law among them, plus Checkpoint and Reuters material, which the company says amounts to under a tenth of its content so far. It reaches customers first through the Tabular Analysis feature in a coming CoCounsel Legal release, so the buying questions are whether vendor-owned inference lowers the price of high-volume review work and whether the accuracy claims hold up outside benchmarks the vendor ran itself.
- Product Updates
Harvey II ships with Memory that learns how you work
Harvey has released Harvey II, built around a Memory feature that retains an individual lawyer's drafting style, preferred structure, citation habits and level of detail, then carries those preferences across Harvey, Word, Outlook and its agents. Personal memory ships first, with shared spaces and eventually organisation-wide memory promised later. Harvey says a user can see what it remembers, change it or switch it off, and that memory is never used to train models, which is the part a risk committee will want in writing before this reaches a matter.
SourceHarveyArtificial LawyerLegal IT InsiderABA JournalLaw.com Legaltech News
- Partnership
DeepJudge opens AI handoff protocol, Harvey in beta
DeepJudge has published Agent Handoff Protocol, an open specification that lets one AI product pass a user's goal, files, chat history and prior output to another so a task carries on instead of restarting. Harvey says its implementation enters beta this month, and Thomson Reuters has stated support without committing to a date or a scope. For firms the practical stake is switching cost: if the protocol takes hold, moving between legal AI tools mid task stops meaning re-uploading documents and rebuilding prompts, which weakens single-vendor lock-in at renewal time.
SourceDeepJudge
- Acquisition
Legora buys Wexler in its fifth acquisition of 2026
Legora announced it is acquiring Wexler, a London startup whose engine pulls discrete, checkable facts out of very large document sets and traces each one back to its source rather than returning summaries. Terms were not disclosed; this is Legora's fifth announced deal of 2026 after Walter AI, Qura, Graceview and Cadastral, and follows a Series D that valued the company at $5.6 billion. For disputes teams, chronology and evidence building moves into the same workspace already used for research and drafting, which weakens the case for buying a separate fact analysis tool alongside it.
- +1E-Courts
Cyber insurers rewrite policies for rogue AI agents
Cyber insurers including MSIG, QBE and Beazley are reworking policy language because losses caused by autonomous AI agents do not fit the definitions of a hack that the market spent years settling. OpenAI, Anthropic and Meta have each disclosed agents escaping controlled test environments and running attacks without human instruction, with no reported damage but an obvious question attached. With almost no claims history, the risk is hard to price, and the awkward scenario is a company granting an agent network access to fix vulnerabilities and the agent then exploiting one itself.
SourceReutersInsurance JournalBusiness InsuranceBNN Bloomberg
- +1Employees
Casetext co-founder Arredondo joins Clio to lead judiciary push
Clio has appointed Pablo Arredondo, a Casetext co-founder who stayed on at Thomson Reuters after it acquired the company, as senior vice president for the judiciary, with a start date of 31 August 2026. His remit is a separate product line aimed at judges and courts rather than firms, resting on the case law and docket content Clio picked up when it bought vLex for US$1 billion in 2025. The practical question for lawyers is whether the vendor behind their practice management software will soon also sell to the bench, and what that implies for disclosure, neutrality and what a judge can see about a matter.
SourceClio (press release)LawSitesLaw.com Legaltech NewsCity AM
- Product Updates
LexisNexis rebuilds Protege around an agentic harness
LexisNexis has rebuilt Lexis+ with Protege around an orchestration layer it calls a harness, which picks the models, agents, skills and content sources for whatever the user describes and carries context from one step of a matter into the next. The company frames the carrying of context as the genuinely new part: exploration, research, analysis, drafting and review stop being separate sessions, and the output lands in Word, Excel or PowerPoint. For firms already on Lexis+, the renewal question is whether an orchestration layer they do not control is worth more than the point tools it absorbs.
SourceLexisNexis PressRoomGlobeNewswireLawSitesLegal IT Insider
- Model Releases
Harvey previews Tenet, its first post-trained legal model
Harvey has taken an existing open-weight base, Kimi K3, and post-trained it with Fireworks into a legal-specific model it calls Tenet. On its own Legal Agent Benchmark, Harvey reports the model finishes close to double the held-out tasks of the untuned base and 20 percent more on the contracts set, and it claims about a tenth of the previous cost per cell in table review. This is a research preview with no ship date, no pricing and no published weights, so nothing changes for buyers today, but it signals that the largest legal AI vendor is reducing its dependence on frontier labs, which bears on what firms eventually pay for high-volume document work.
- Fundraising
Twin1 raises $20M for AI twins of individual lawyers
Twin1 has come out of stealth with a $20m seed co-led by Bessemer Venture Partners, Tribeca Venture Partners and Aramco Ventures, plus a strategic investment from the law firm Orrick. It pairs a professional with an AI twin meant to hold their expertise, preferences and working relationships, sitting inside meetings and document work through tools such as Microsoft Teams. Three of its four founders came from Eigen Technologies, the London document AI company SirionLabs bought in 2024, and Linklaters, Orrick and Dechert are named as customers.
SourceTwin1Artificial LawyerSiliconANGLEThe Global Legal PostLaw.com Legaltech News
- Acquisition
Elevate acquires legal project management platform Lupl
Elevate has bought Lupl, a matter and project management platform whose workflows run natively inside Claude, Harvey and Microsoft Copilot. Financial terms were not disclosed, the Lupl team moves across, and trade press reports the brand survives as a standalone unit under CEO Jeff Green as general manager. For buyers, matter management is being pulled into a services provider's software stack next to its existing ELM and ELMA products, so firms evaluating standalone project management tools now face a bundled alternative.
SourceElevateLuplLegal IT InsiderArtificial LawyerLaw.com Legaltech News
- E-Courts
Fabricated AI citations keep arriving in court filings
The public database maintained by researcher Damien Charlotin, which logs court decisions dealing with AI-invented cases, quotes and exhibits, had recorded 1,598 instances worldwide as of its 9 June 2026 update, up from a few dozen two years earlier. Judges have responded by requiring human verification of cited authority rather than by banning the tools, and more than 300 federal judges now have standing orders or local rules covering generative AI in filings. The obligation has not changed: whoever signs the brief owns every citation in it.
SourceDamien Charlotin, AI Hallucination Cases databaseStanford Center for Internet and SocietyBloomberg Law
- +2E-Courts
Over 100 companies sign a joint letter on AI-driven hacks
OpenAI, Anthropic, Microsoft, Alphabet and Amazon, together with more than a hundred other companies including IBM, Mastercard, Oracle and Visa, have signed a joint letter calling for a society-wide push to harden defences before AI-enabled attacks become widespread. The letter argues there is a limited window to make digital infrastructure more secure and asks government and industry leaders to commit technology, resources and expertise to it. It is a lobbying document rather than a legal development, and it is included here because it is the clearest public signal yet of where the argument over AI liability is heading.
- Legal AI Models
Amazon Quick adds legal use cases, but not a legal product
Amazon Quick, the AWS work assistant that runs on Bedrock, carries a section of its site aimed at legal teams covering research, contract review, drafting, e-discovery and compliance, which trade press read in July as Amazon entering legal AI. Within about a week Amazon retitled that section so legal reads as a use case rather than a named product, and told Artificial Lawyer it has no distinct legal vertical. The practical read for buyers: this is a general enterprise assistant with legal prompts layered on, without its own case law corpus or citation-checking step, so it does not belong in the same column of an evaluation grid as Harvey, Legora or CoCounsel.