Here’s a quick look at the biggest AI news from the month. We’ve pulled together the headlines shaping technology, business, and policy.
The EU AI Act's August 2 milestone arrives, and enforcement powers switch on
August 2, 2026 was the date every AI Act compliance roadmap was built around, and it has now passed with a narrower set of obligations live than originally planned. As of this date, the AI Office and national authorities are responsible for implementing, supervising, and enforcing the Act, and the AI Office's enforcement powers over general-purpose AI models take effect, carrying fines of up to 15 million euros or 3% of a company’s global turnover. The transparency obligations in Article 50, including AI-interaction disclosure, also apply now, while the grace period for labeling AI-generated content runs to December 2, 2026. The high-risk obligations that were once due today have been formally deferred: stand-alone Annex III systems (employment decisions, profiling, access to essential services) now apply from December 2, 2027, and high-risk AI embedded in regulated products from August 2, 2028. Those deferrals are no longer provisional. The Digital Omnibus on AI was formally adopted after the Parliament's June 16 vote and the Council's June 29 sign-off. For legal ops teams at global companies, the practical move is to separate what is enforceable now (transparency, GPAI enforcement) from what you have runway on (high-risk classification). [European Commission] [AccuroAI]
AI hallucination sanctions keep climbing, and now they include suspensions
The running tally of court decisions involving AI-fabricated citations is approaching 1,500 worldwide, with more than 1,000 in the United States, according to the database maintained by researcher Damien Charlotin. Penalties have escalated well beyond the four-figure fines of the early cases, reaching $15,000 per attorney at the federal appellate level and, for the first time, bar suspensions tied to AI filings. The pattern across the cases is consistent: the cover-up draws the harsher penalty, and candor with the court is the one thing an AI cannot do for the lawyer who relied on it. For legal ops teams, this reinforces the lesson from the Sullivan & Cromwell matter earlier this year. Verification is now a professional duty, and review processes need to be redesigned for output that looks finished when it is not. [ABA Journal] [Damien Charlotin's database].
The US federal-versus-state fight over AI regulation grinds on
There is still no comprehensive federal AI statute, and the standoff between Washington and the states continues to define the compliance picture. The Trump administration's December 2025 executive order set a "minimally burdensome" national policy and stood up a DOJ AI Litigation Task Force to challenge state laws, which then intervened in xAI's constitutional challenge to Colorado's AI Act. At the state level, momentum has not stopped: New York's AI moratorium bill is awaiting action from Governor Hochul, Maine's was vetoed, and consumer-protection and disclosure bills continue to advance. For legal ops professionals advising on AI governance, the fifty-state patchwork is not going away, and tracking state-by-state developments remains a core part of the job. [IAPP] [TechPolicy.Press]
"Shadow AI" is the governance gap legal ops keeps underestimating
Thomson Reuters' 2026 Future of Professionals report found that 34% of professionals use AI tools their organization has not sanctioned, in ways it cannot see. The same research found that firms with a named AI strategy see 66% of professionals reporting that AI meets or exceeds expectations for creating value, compared with just 22% at firms without one. The risk for legal teams is not only which tools get selected, but the ungoverned use happening around whatever is officially approved, which carries real confidentiality and privilege exposure. For legal ops teams, the actionable move is to get ahead of shadow adoption with a clear policy and sanctioned tooling that people actually want to use, rather than a ban that pushes usage further out of view. [Thomson Reuters]
Agentic AI adoption keeps outrunning governance, and the gap is now the story
Agentic systems are moving from pilots into everyday legal workflows, visible across CoCounsel's agentic rebuild, Harvey's agent expansion, and vendor benchmarks for agentic legal reasoning. Gartner still projects that as many as 40% of enterprise applications will incorporate task-specific AI agents by the end of 2026, and regulators remain cautious, with the EU flagging agentic capabilities in its systemic-risk considerations for general-purpose models and the Bar Standards Board advising barristers to treat agentic AI as high risk. The governance challenge is different in kind from traditional AI oversight: an agent that performs well today may not in three months, as models drift, upstream data changes, and business processes evolve into edge cases the agent was never designed for. For legal ops teams, that means governance has to be continuous and monitored, not a one-time approval at procurement. [Gartner] [Bar Standards Board]
That's the roundup for this edition. As always, these stories are curated with the legal ops professional in mind. Join the conversation in the OnPoint Community AI Peer Group
