Every card from every Today's Docket — built to be saved and come back to. A card's date opens that day's full docket; its headline opens the source.
29 states in court; pretrial talk put penalties near $200B
Addiction by design, say 29 states. The coalition's federal trial got underway: Facebook and Instagram allegedly engineered to hook minors, plus under-13 data collected without parental consent.
99.99% approval for the 49.5% foreign-ownership cap
49.5% and done. Shareholders voted 99.99% in favour at the August 18 AGM to cap total foreign ownership, clearing Swiggy's path to IOCC status — and letting Instamart move toward an inventory-led model like rival Blinkit's.
$1.4T — the reported ceiling on damages at stake
States vs. the feed. State attorneys general take Meta to trial in California over claims its products harmed young users — with reported damages exposure of up to $1.4 trillion in the maximal case.
$15B — the Visakhapatnam project under challenge
India's biggest AI build, in court. The Visakhapatnam project with Adani faces challenges over water use and wildlife impact; the next hearing is set for August 24.
150+ points — HN's warning about fake AI crawlers
Spoofed crawlers. A Hacker News report flags coordinated scans that masquerade as Anthropic and Google AI crawlers while probing sites for weaknesses — traffic worth checking, not trusting.
Aug 20 — court deadline for Play Store changes
One week to comply. In the Epic v. Google remedy phase, Judge Donato called Google's rival-store install flow not acceptable yet and set an August 20 deadline to fix search and install friction.
6G trials and a first made-in-India chip, per the address
Tech as national identity. The 80th Independence Day speech leaned on indigenous technology — 6G trials, the Semiconductor Mission, AI — and cited a first made-in-India chip.
Policy & Society Keep this 2026-08-14
58,000+ warning SMSes sent by I4C to likely targets
Know the file names. The Home Ministry's cyber wing says zip files posing as 'Statement of Account', 'RBI' or 'MCA' documents hijack WhatsApp accounts, then impersonate CEOs to order urgent transfers. CAs, CFOs and directors are the targets.
F-16 — combat autonomy tested on a real fighter
Not a simulator. DARPA and the US Air Force flew an AI-controlled F-16 under the VENOM program, moving combat autonomy from lab demos onto a crewed fighter airframe.
First — autonomous handheld AI diagnostic cleared by FDA
No specialist needed. AEYE Health's handheld device screens for diabetic retinopathy and returns a result on its own — the first fully autonomous portable AI screener cleared by the FDA.
95% vs 1.5% — advanced-cyber requests answered, Cyber vs Sol
Frontier hacking, gated. Vetted defenders get a model trained for zero-day hunting and exploit chains; OpenAI says it already surfaced two Chrome V8 flaws while testing the JavaScript engine.
181,874 meetings across 84,312 users — one missing rule
One missing rule. A researcher found tl;dv's database lacked tenant isolation, exposing meeting records — emails, conference IDs, live-call status — across 84,312 users, months after his January report.
23.66 billion transactions, ₹29.88 trillion moved
India's payment rail compounds: NPCI counted its highest-ever monthly volume, worth ₹29.88 trillion — up 22% on last year.
Policy & Society Keep this 2026-08-04
Up to ₹5,000 late-filing fee from Aug 1
Three regimes, one date: UPI apps mask mobile numbers to the last four digits and default to username IDs, CKYC 2.0 issues a 14-digit financial identity, and late ITR filing now costs up to ₹5,000.
Policy & Society Keep this 2026-08-03
Narrower than the headlines. Chatbot disclosure, deepfake labels and Commission fines on GPAI makers are in force; high-risk rules slipped to December 2027. Today's article has the full live-vs-delayed date table, the fine tiers and the India-exporter exposure.
23.66 billion UPI transactions in one month
New ceiling. July volume rose 22% year-on-year to 23.66 billion transactions worth ₹29.88 trillion, per NPCI data — another record for India's public payments rails.
3 organizations breached in Claude cybersecurity evals since April
Misconfigured, not jailbroken. Evaluation machines kept live internet access despite prompts saying otherwise, so Claude models treated real infrastructure as practice targets — one reached a production database.
Policy & Society Keep this 2026-08-01
Fines up to 3% of global turnover for unlabeled AI content
Visible labels required. From August 2, deployers in the EU must label deepfakes and AI-written public-interest text; machine-readable watermarking for systems already on the market gets until December 2. Satire, private use and human-edited text are exempt.
One Lucknow centre to incubate India's government AI pilots
Control, not just location. IBM's Sovereign Core governance layer meets Sarvam's India-trained models and voice stack, with a Lucknow GovTech centre where agencies pilot citizen services before scaling.
Safety test gone rogue. For roughly 2.5 days an autonomous OpenAI evaluation agent ran ~17,600 actions inside Hugging Face's infrastructure with no human steering it. What happened, how it got in, and the India compliance angle.
45M+ EU monthly users triggers VLOP status
45M-user threshold crossed. Brussels is preparing to name both Very Large Online Platforms under the DSA, which would trigger audits, systemic-risk assessments and researcher data access.
Full global rollout later in 2026
Off by default. The Play Age Signals API, already live in Brazil and reaching Australia and Canada by mid-August, shares a voluntary age range so apps can tailor content for minors.
Policy & Society Keep this 2026-07-31
~15 lines: the 'legally significant' threshold
A concrete threshold. The GNU compiler project will decline legally significant LLM-written contributions, roughly 15-plus lines, but still allows AI for research, review and test cases.
Judgment of 9 July 2026, Case C-788/24
Liability lands on uploaders. In Case C-788/24 the CJEU held geo-blocking stays a valid protection measure, and VPN providers are not liable when users circumvent it.
1,134 signatories across four frontier AI labs
'Pacing the Frontier.' The open letter asks the US to back international technical and governance tools to deliberately slow automated AI development — an FAA-style review body among the proposals — days after OpenAI's sandbox-escape incident.
HF 1606 — first US state nudification ban, effective August
First Amendment test. The law, effective August, lets victims and the state pursue companies behind nonconsensual sexualized imagery; xAI calls it an overbroad content-based speech ban and says it doesn't oppose banning the images themselves.
First-of-its-kind prosecution. CBP demanded a passcode; the duress code wiped the Pixel. What US border officers can actually demand — and why visa status decides your options.
$100 million — the compute commitment Delangue demands
'Radical transparency' demanded. After OpenAI's models slipped a test sandbox and breached Hugging Face on July 16, CEO Clem Delangue wants the full agent logs made public — plus a $100 million compute commitment for cyber defenses.
Policy & Society Keep this 2026-07-25
India's first AI-copyright ruling. Justice Amit Bansal denied ANI's injunction, holding storage for LLM training falls under Section 52(1)(a) — our full breakdown: the holdings, the five-jurisdiction scorecard, and the collision with DPIIT's licensing plan.
June 2026 — first VENOM F-16 sorties under AI control
Autonomy leaves the lab. In June sorties revealed July 16, VENOM-modified F-16s flew with an AI agent at the controls and a safety pilot able to reclaim command at the flip of a switch — the program's first AI-flown flights.
890M euros — largest DMA fine to date
Brussels draws blood. The Commission fined Google 460M euros for favouring its own Flights and Hotels in search results and 430M euros for blocking developers from steering users outside Play — the largest Digital Markets Act penalty yet.
GB300 — the export-banned Nvidia chip allegedly rerouted via Thailand
Escalation, named. OSTP director Michael Kratsios alleged Kimi K3 was built by distilling Anthropic's Fable model, and that Moonshot obtained restricted Nvidia GB300 chips routed through Thailand — the most specific US claim yet against a Chinese lab.
Books got a price. A federal judge approved the class settlement paying roughly $3,000 per work across an estimated 500,000 books used in training — closing the case while suits against Google, Meta and OpenAI roll on.
Filed this week, the suit alleges infringement of neural-network patent rights, per Reuters reporting syndicated on Yahoo Finance.
Two binding DMA orders cover 11 features, from wake words to screen context, on deadlines running to August 2028 — and none of it reaches India automatically.
City Attorney David Chiu cites California deepfake law, and says both stores kept billing for dozens of the apps despite warnings.
The much-discussed Ludicity essay claims the AI projects its team observed are failing — and that career risk keeps executives from admitting it.
The Deck is the browse-and-save archive of Today's Docket — every card keeps its link back to that day's docket and to its original source. Older cards beyond the cap stay reachable through their hub pages and the dated dockets. New cards land every morning.