Here’s a quick look at the biggest AI news from September 2026. We’ve pulled together the headlines shaping technology, business, and policy.
California becomes the first state to write attorney AI duties into statute
On September 30, Governor Newsom signed SB 574, making California the first state to put rules for attorneys' use of generative AI directly into statute rather than leaving them to ethics guidance. The Legislature had passed the bill unanimously on August 31, and it takes effect January 1, 2027. The core obligations will look familiar to anyone tracking the hallucination cases: an attorney cannot delegate the practice of law to generative AI, cannot enter confidential or nonpublic information into a system that does not restrict access, must take reasonable steps to verify AI output and correct errors, and must personally verify every citation in a filing, including ones the AI supplied. Filers also have to disclose AI use to the court for documents submitted to it. Several commentators note the law largely codifies duties that already existed under the Rules of Professional Conduct, which is the point: it gives a court and the State Bar a second statute to point to, one that names AI specifically. For legal ops teams, this is the moment to turn your AI usage policy into something auditable. If your attorneys are in California, the verification and confidentiality steps now carry statutory weight, and your tooling and intake processes should make compliance the path of least resistance. [Bloomberg Law] [Sullivan & Cromwell]
State AI lawmaking accelerated in September while the federal-versus-state standoff held
SB 574 was one piece of a much larger month in Sacramento. Newsom closed out the signing period on September 30 by enacting another batch of AI bills, capping a 2026 session that saw nearly 30 AI-related measures become law in California alone. The headline from that final batch is the No Robo Bosses Act (SB 947), which bars employers from relying solely on automated systems to fire or discipline workers and requires a human reviewer to corroborate those decisions. Earlier in the month, Newsom signed an executive order pushing for independent oversight of AI companies and the potential requirement of an emergency shutoff, or "kill switch," for advanced models, explicitly framing it as filling a gap left by Washington. Illinois and Oregon governors took parallel executive action the same month. The federal direction still runs the other way, with the administration's deregulatory posture and its task force challenging state AI laws. For legal ops professionals, the fifty-state patchwork is not just intact, it is thickening, and the governance tracker you maintain needs to cover employment AI and frontier-safety rules, not only disclosure and bias law. [CNBC] [Washington Post] [Transparency Coalition]
The hallucination case count crossed 2,000, and the penalties kept climbing
The tally this roundup put near 1,500 last month has kept its pace. Damien Charlotin's AI Hallucination Cases database had logged more than 2,000 court decisions worldwide involving AI-fabricated material by late September, reaching 2,125 as of October 2. The curve tells the story: roughly 200 cases in mid-2025, about 1,600 by June, and past 2,000 by September. Penalties have moved well beyond the early four-figure fines, with the largest single-matter sanction now around $109,700 in a District of Oregon case, per the ABA Journal. The lesson from the earlier high-profile matters has not changed: verification is a professional duty, not a nice-to-have, and the cover-up tends to draw a harsher penalty than the original error. For legal ops teams, this reinforces why SB 574's verification requirement is less a new burden than a formalization of what review processes should already do. Build the citation check into the workflow so it happens before a filing goes out, not after a judge flags it. [Damien Charlotin's database] [AI Hallucination Cases Tracker]
Adoption is now the baseline, and the governance gap is the open problem
A report from the Association of Corporate Counsel found that more than four out of five legal departments have now adopted AI, but many still lag on the rules to govern it and the metrics to measure whether it is working. A separate September report pointed the same direction, with e-discovery company Disco finding that 58% of legal professionals have integrated AI into their routine processes. Adoption, in other words, is no longer the story. The gap between how many teams use AI and how many can govern or measure it is. For legal ops professionals, this is the opening. Governance is increasingly the work that legal ops owns, and the teams that can show a named policy, defined ownership, and a way to measure value are the ones turning AI from a line item into a demonstrated return. For more information, read our State of AI in Legal: 2026 Survey Results. [Law360 Pulse] [ABA Journal]
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
