August 2026 legal tech recap for in-house teams: EU AI Act transparency, Thomson 1.0 and Harvey II, Gemini for Legal, ILTACON and California's SB 574.

Legal Tech Rewind is our month-by-month look back at 2026 for in-house legal teams: what happened, why it mattered, and what it looks like with the benefit of hindsight. Published October 2026.
The EU's AI transparency rules went live, the legal vendors decided they wanted their own AI models, and everyone at ILTACON agreed that the future is tools that talk to each other.
On August 2, the EU AI Act's Article 50 transparency obligations started to apply. Companies must tell people when they're interacting with AI, label deepfakes and certain AI-generated content, and mark generative outputs in a machine-readable way. Generative AI systems already on the market have until December 2, 2026 for marking. Fines run up to €15 million or 3% of worldwide turnover. Meanwhile, the high-risk deadlines were already pushed to December 2027 and August 2028 by the Digital Omnibus, which entered into force on July 27.
Anthropic responded by saying Claude outputs would carry invisible machine-readable marking worldwide, not just in the EU. Artificial Lawyer noted that metadata can be stripped and that some uses, like proofreading or translation, may not carry a watermark.
Why it mattered: if your company runs a customer-facing chatbot or publishes AI-generated content that reaches the EU, this isn't a 2027 problem. It's a now problem.
What to do: confirm your chatbot disclosures and content labeling, and check whether documents your team drafts with AI now carry detectable markers.
Why it mattered: Thomson Reuters had announced its model plans back in March, but after June's model shutdown, the case for controlling more of the stack got much easier to make. For buyers, it's one more question for the vendor review: which model runs my work, and can that change without telling me?
Watch out for Memory: when your outside counsel's AI "remembers" lawyer preferences and shares spaces with clients, ask what it retains about your matters. That belongs in your outside counsel guidelines.
Google launched Gemini Enterprise for Legal in preview on August 25, with skills for contract lifecycle work, citation verification, regulatory horizon scanning, data subject access requests and NDA drafting, plus connectors to Docusign, iManage, NetDocuments, Relativity, Thomson Reuters, Harvey and Legora. Launch customers were all law firms. Anthropic hired Robert Mahari as its first Head of Claude for Legal.
Why it mattered: if your company runs on Google Workspace or Gemini Enterprise, legal-specific AI may arrive through a license IT already manages, which makes it a governance question as much as a buying one.
ILTACON in Nashville drew 5,700 registrations, up from 4,600 in 2025. LawSites' Bob Ambrogi described "conspicuous spending" by Harvey and Legora and asked whether we'd reached "peak legal tech," predicting a correction or realignment rather than a burst bubble. The dominant theme was integration: meet lawyers where they already work.
Two announcements made that concrete:
Why it mattered: a multi-vendor AI strategy becomes realistic when tools can hand work to each other. Add "do you support MCP and Agent Handoff Protocol?" to your RFP template.
BigHand bought AI pricing specialist Ayora, Anaqua bought Unified Patents, and Elevate bought legal project management platform Lupl, which it plans to integrate with its tools for law departments. Artificial Lawyer asked, "Is This The Great Legal Tech Consolidation?" Harvey was reported to be in talks to raise $500 million. (It closed $550 million in September.)
Why it mattered: when a vendor in your stack gets acquired, the roadmap you bought can change. Check your change-of-control, data portability and price protection clauses now. Our vendor acquisition checklist walks through it.
LegalOn's 2026 In-house Pulse found 92% of in-house teams use AI, but 54.9% are still experimenting without shared workflows, only 8.8% ground AI in their own playbooks, and just 2% have full workflow support from intake through post-signature. LegalOn is a vendor and didn't state the sample size, so treat it as directional.
Why it mattered: most in-house teams use AI one person at a time. The value comes when the prompts and playbooks are shared.
On August 31, California's legislature unanimously passed SB 574. It requires lawyers to verify AI output and every citation, keep confidential information out of non-secured AI systems, and disclose AI use to courts, and it states that a lawyer "shall not delegate the practice of law to generative artificial intelligence." Gov. Newsom signed it on September 30.
Also in August: the LeDoux v. Outliers case ended with claims dismissed with prejudice after the court excluded an expert whose report cited AI-hallucinated articles.
Why it mattered: in-house lawyers licensed in California are covered too. Your AI policy needs verification, confidentiality and court-disclosure steps.
August's lesson: everyone wants to own the model. Your job is to own the data, the playbook and the off switch.
More from the 2026 Legal Tech Rewind: ← July 2026 · September 2026 → · The full 2026 Rewind: 9 lessons
This article is general information, not legal advice.