Thelonious ICYMI — Edition 19 — 13 August 2026
Thelonious ICYMI — Edition 19 — 13 August 2026
Welcome to the 19th edition of ICYMI — your weekly briefing on the developments shaping the AI legal and regulatory landscape, curated from Thelonious.
A week in which accountability kept moving up the chain, but sideways too. Pennsylvania went after a design choice. Congress went after a procurement choice. A press alliance went after a broken promise. Read together, the question is no longer only who built the harmful system — it’s who chose it, who signed off on it, and who profits when it costs someone else a job.
Here’s a snapshot of what’s inside this week:
- Pennsylvania v. TikTok — The Commonwealth of Pennsylvania alleges TikTok violated the state’s Unfair Trade Practices and Consumer Protection Law, claiming the platform secured a 13+ age rating by misrepresenting its own content while intentionally engineering addictive features it knew were harmful to young users’ mental health. The allegations are untested. What distinguishes this from Tennessee’s case against Meta, covered in Edition 17, is the statute: this is pled as a consumer protection claim, not fraud or breach of fiduciary duty, and nearly every US state carries some version of an unfair trade practices law on its books. A ruling that algorithmic design choices can constitute a deceptive practice would hand every state attorney general a template that reaches well beyond TikTok, and well beyond social media.
- Congressional Investigation into DoorDash — The House Committee on Homeland Security and the House Select Committee on the Chinese Communist Party have opened an inquiry into DoorDash’s use of an AI model, Kimi K2.6, for internal code evaluation, centred on alleged ties between the model’s developer, Moonshot AI, and China’s military. No findings have been made. The significance sits less in DoorDash’s specific exposure than in what the inquiry represents: legislative scrutiny is shifting from what a company’s AI systems output to which vendor built the underlying model. Procurement decisions for internal tooling — which model handles code review, which model handles customer service — are being folded into the same national security due diligence regime that already governs hardware supply chains and foreign ownership review.
- APIG v. Google — The Alliance de la Presse d’Information Générale alleges that Google’s AI Overviews and AI Mode features used French press publishers’ content without prior authorisation, good-faith negotiation, or compensation, in breach of commitments Google made to the French Competition Authority in 2022. Google has not conceded the claims. The prior commitments carry the case: APIG is not arguing Google should have sought permission, it is arguing Google already promised to and didn’t, converting what would otherwise be a scraping dispute into a breach-of-settlement dispute with a paper trail already on file with a regulator.
- California Announces AI Cyber Defense Program — Governor Gavin Newsom launched the AI Cyber Defense Program on 10 August 2026, a first-in-the-nation initiative using AI tools to identify and mitigate AI-enabled threats to the state’s critical infrastructure. Set against the UK’s Cyber Shield initiative from Edition 14, the program confirms that defensive, agentic AI deployment is spreading from national governments to individual US states, ahead of any comparable federal framework.
- AI Tax and Work Protection Act — H.R. 10044 would impose a federal excise tax on the use of large foundation models by “covered persons” who develop, sell access to, or modify those models to generate revenue or reduce their workforce. The workforce-reduction trigger is the operative mechanism: the bill is drafted to tax automation-driven job cuts specifically, rather than AI revenue generally, placing it in the same legislative lineage as Senator Markey’s AI Accountability Agenda from Edition 15.