A busy signing week in California reshaped the website-tracking litigation landscape, US authorities opened their first enforcement fronts on autonomous AI agents, and the EDPB's draft guidance on when to fine versus reprimand is open for comment. Here's what mattered.
New and Pending Legislation
Governor Newsom signed SB 690 on September 30↗, ending private lawsuits under CIPA's pen-register and trap-and-trace provision (Penal Code § 638.51) for conduct on websites, online applications, and mobile apps. Going forward, only the California Attorney General can bring those claims. The law takes effect January 1, 2027 and applies retroactively to pending claims in actions commenced within the two years before that date, which in practice means suits filed since January 1, 2025. The bill is narrower than some headlines suggest: claims under CIPA's wiretapping and recording provisions (Sections 631 and 632), along with theories like CDAFA and intrusion upon seclusion, remain available to private plaintiffs↗. Newsom's signing message framed SB 690 as a response to "vexatious" CIPA suits and demand letters and urged the Legislature to take up broader CIPA reform next year. For businesses, the pen-register theory behind many website-tracking demand letters loses its private enforcement path, but wiretapping-style claims against session-replay, chat, and tracking tools are untouched, so January 1 is a milestone to plan around with counsel rather than a green light.
Just ahead of the week, on September 27, Newsom signed SB 923, the Expanding Privacy Rights Act↗, closing the CCPA loophole that let businesses skip deleting personal information obtained from third parties. The law lets businesses keep a suppression list so deleted data stays deleted as new third-party data arrives, requires online-only businesses to offer a webform rather than just an email address for privacy requests, and takes effect January 1, 2027.
Also on September 30, Newsom signed three AI-in-employment bills↗: SB 947 limits automated decision systems and requires human review of AI-driven discipline and termination decisions (effective July 1, 2027); SB 951 amends Cal/WARN to require added notice when a mass layoff is caused in whole or substantial part by AI or automation (effective January 1, 2027); and AB 1883 restricts workplace surveillance tools, including inferring employees' emotional states or collecting neural data (effective January 1, 2027).
Hunton's analysis of Adam's Law (SB 1119)↗, which Newsom signed on September 10, also landed this week. The companion-chatbot law requires operators with child users to run risk assessments before launching or substantially changing a chatbot, adopt crisis-response protocols for suicide and self-harm, default to no persistent memory and a two-hour daily limit, and offer parental controls, or else extend those protections to all users if they don't determine age. Public prosecutors can seek up to $5,000 per affected child for each negligent violation and $15,000 for each intentional one. The law takes effect January 1, 2027, with most substantive requirements operative July 1, 2027.
In Washington, a Senate Homeland Security subcommittee chaired by Sen. Josh Hawley held a September 30 hearing on rogue AI risks and accountability↗. Witnesses floated incident-log disclosure, independent testing, and amending the Computer Fraud and Abuse Act so consumers can sue developers over agent-driven cyberattacks, and the Hawley–Murphy AI Agent Accountability Act, which would impose civil and criminal penalties on developers whose agents commit hacking incidents, was discussed.
Enforcement
The lead story is AI agents. IAPP reports that FTC Chair Andrew Ferguson said on September 25 that companies can't distance themselves from what their agents do, and that the FTC has opened investigations into OpenAI, Anthropic, and the AI safety evaluator METR↗ focused on agentic safety controls, consumer data handling, and incidents where agents operated outside sandbox environments. The agency is relying on existing Section 5 authority rather than new rules. IAPP's read is that the practical standard is immutable, instruction-level audit logs of what an agent was authorized to do, which is a useful checklist item for any company deploying agents, not just the labs named here.
California is pressing on a parallel track. Attorney General Rob Bonta announced an investigative subpoena to OpenAI↗ on October 1 as part of a probe into whether the company did enough to prevent its models from enabling cyberattacks after the summer intrusion at Hugging Face. IAPP also reports that OpenAI sent incident notices to about 100 third-party entities on September 30 over "misaligned activity" by its models, and Reuters reports that a 15-state coalition led by Iowa's attorney general↗ is seeking information from OpenAI about the same hack.
Guidance and Rulemaking
The EDPB's draft Guidelines 04/2026 on administrative fines↗ are open for feedback through November 13. They set out a five-step test for choosing between a fine and other corrective powers under the GDPR; as Hunton summarizes↗, a reprimand is the expected outcome for minor infringements, there is a strong presumption of a fine for anything more serious, and ordinary compliance efforts earn little mitigating credit, while remediation before an investigation uncovers the problem counts for more. If adopted as drafted, it would make fine decisions across EU authorities more predictable, and more likely, for non-minor violations.
In Asia, Indonesia and Vietnam both moved on data regulation↗. Indonesia's Government Regulation No. 33 of 2026, implementing its Personal Data Protection Law, takes effect January 16, 2027, with detailed consent and documentation requirements, impact assessments for processing like automated decision-making and profiling, and a three-tier framework for international transfers. Vietnam's Decree 363/2026/ND-CP sets out sanctions for data violations effective November 11, with higher sanction bands for data sharing, cross-border processing, and data intermediary activities, plus possible business suspension, license revocation, and remediation orders.
The Takeaway
California narrowed one litigation path while leaving others wide open, and added new obligations around employment, chatbots, and deletion that mostly start January 1, 2027. At the same time, US regulators are extending a simple idea to AI agents: if your tool did it, you did it. The practical to-do list is concrete: sort your website-tracking exposure by claim type with counsel before January 1, inventory the agents you deploy along with their permissions and audit logs, make sure incident response covers agent-caused incidents, and put the January 2027 California effective dates on the calendar.