> daily_signal(2026_07_22)
A court just priced a pirated training book at $3,000. DHS is tripling its AI border towers by 2034. Apple started recording the Genius Bar. Samsung's new assistant reads your health data back to you.
PickBits Daily Signal · Wednesday, July 22, 2026
1. Continuing 07-07 #3: with Anthropic's $1.5 billion book-piracy settlement now finalized by a federal judge on July 20, 91% of the 482,000 covered books have already been claimed for payment — proof the roughly $3,000-per-work payout is real money moving, not a projection, while the ~100 authors who opted out are still betting they can beat it in court.
The new number is not $1.5 billion and not even $3,000 — both of those we already reported on 07-07. It's 91%: the share of the settlement's 482,000 covered books already claimed by rights-holders, now that Judge Araceli Martínez-Olguín granted final approval on July 20. That number didn't exist three weeks ago, because there was nothing yet to claim. It's the first real evidence that a court-priced piracy settlement actually disburses at scale, not just on paper. The underlying legal split still matters for readers tracking AI copyright cases generally, though it isn't new to this approval: the court did not rule that training on copyrighted text is illegal — Anthropic's fair-use position on training survived. What it could not survive was acquisition. Anthropic built its library two ways: books it bought and scanned, which was fine, and books it pulled from Library Genesis and Pirate Library Mirror, which was not. That holding was already true before July 20; what changed on July 20 is that the payout stopped being theoretical.
Key fact: IF YOU WROTE OR PUBLISHED A BOOK THAT MAY BE IN THE CLASS, CHECK WHETHER YOUR TITLE IS COVERED AND FILE A CLAIM — about 9% of the more than 482,000 covered books were still unclaimed as of approval, and each claimed work carries roughly $3,000. The class administrator's works list and claim process are linked from the court-approved settlement notice in Bartz v. Anthropic (N.D. Cal.); start from the case coverage at https://techcrunch.com/2026/07/20/anthropics-landmark-1-5b-copyright-settlement-is-approved/ and follow through to the official claims site rather than any emailed link, because settlements of this size reliably attract phishing.
TechCrunch (2026-07-20): Judge Araceli Martinez-Olguin, U.S. District Court for the Northern District of California, approved the settlement Monday, July 20, 2026 — 'The payout will deliver $3,000 per work across an estimated 500,000 works, shared among the authors and publishers who hold rights to them.' · TechCrunch: Anthropic admitted it 'illegally downloaded and stored millions of copyrighted books' via Library Genesis and Pirate Library Mirror, while its position that TRAINING on copyrighted text is fair use was upheld — liability attached to acquisition, not to training · AP (via US News, 2026-07-21): the judge said the class-action settlement provides 'meaningful relief'; about 91% of the more than 482,000 books covered have been claimed by authors or publishers now due payment · AP: plaintiff attorney Justin Nelson called it 'the largest known copyright recovery in history'; the suit was brought in 2024 by thriller novelist Andrea Bartz with two other authors · primary source
2. DHS plans to nearly triple its network of AI surveillance towers along US borders — from 830 today to 2,300 by 2034, at a cost of $1 billion in already-appropriated money — according to a Government Accountability Office report the Electronic Frontier Foundation broke down on July 20.
These are not cameras someone watches. The Integrated Surveillance Tower program's newer builds — including Anduril Industries' Sentry model — combine radar, thermal infrared and optical systems with AI-based software that autonomously detects, classifies and tracks what it decides is a target, over long distances, and hands a human the alert rather than the raw feed. The consequential shift is therefore not the count going from 830 to 2,300; it is that the judgment about what counts as suspicious movement is being made by a model, at scale, on a permanent installed base, in places where the people being classified are overwhelmingly not crossing anything. EFF's read of the GAO findings puts the buildout at $1 billion drawn from the 2025 tax-and-spending law, with deployment concentrated on the US-Mexico border and specific expansion into Arizona, Texas, and California coastal communities — which is to say into towns where residents, not just crossers, live inside the detection cone. The accountability gap is the part worth flagging to anyone who works in this space: the same GAO annual assessment noted the tower program did not hit major fiscal-2025 milestones, so this is an expanding, largely unevaluated AI system whose false positives are borne by people with the least practical ability to contest them.
Key fact: IF YOU LIVE IN OR NEAR A BORDER COMMUNITY IN ARIZONA, TEXAS, OR COASTAL CALIFORNIA, FIND OUT WHAT IS ALREADY NEAR YOU BEFORE THE NEXT TRANCHE LANDS — EFF maintains a public map and dataset of existing border surveillance towers as part of its Atlas of Surveillance work, and the tower locations are the input to any local comment or council fight. Start at https://www.eff.org/deeplinks/2026/07/explosion-surveillance-towers-coming-us-borders-costing-over-1-billion
EFF Deeplinks (2026-07-20), reading a July 2026 GAO report: DHS plans to grow surveillance towers along US borders from 830 to 2,300 by 2034 — 'DHS expects to expend $1 billion in taxpayer dollars for this dangerous expansion' · EFF: the $1 billion 'comes from the so-called One Big Beautiful Act—a massive tax and spending law that President Trump signed in 2025' · EFF: the Integrated Surveillance Tower program includes autonomous towers with 'radar, thermal infrared and optical systems to track targets over long distances,' using AI-based systems for autonomous target tracking; Anduril Industries is named as the provider of the Sentry autonomous tower model · EFF: deployment is concentrated on the US-Mexico border, with named expansion into Arizona, Texas, and California coastal communities · GAO-26-108118 (DHS Annual Assessment): Integrated Surveillance Towers did not achieve major milestones in fiscal year 2025, with milestones scheduled for fiscal year 2026 · primary source
3. Apple has started recording and AI-summarizing conversations between Genius Bar employees and customers at a limited number of stores under a pilot called Live Notes, Bloomberg's Mark Gurman reported July 19 — both the worker and the customer must opt in, and workers still say the tool could be turned into a monitoring system.
Strip the Apple branding and this is the template every frontline service job is about to meet: an AI listener that transcribes the customer interaction, generates a summary, and files it to the work record. Apple's design choices are, on paper, the careful ones — Live Notes only activates if both the employee and the customer consent, Apple says the audio recordings themselves are not saved, managers do not receive the transcripts, and the employee can edit both the transcript and the summary before it is saved into the internal Genius Bar system. The stated purpose is documentation time, which is a real burden on Genius Bar staff. What makes this a story rather than a product note is the gap between the current controls and the structural capability: once the transcription and summarization pipeline exists and is wired into the work record, the constraint on using it for performance evaluation is policy, not architecture. Apple Store workers quoted in the coverage say exactly that — the concern is not what Live Notes does today, it is that 'something almost exactly like this could be used for monitoring and evaluating employees' tomorrow, and the policy that currently forbids it is a sentence Apple can rewrite. For anyone who owns a service-desk process, this is the decision you will be asked to make within the year, with the pilot's consent model as your best available reference.
Key fact: IF YOU MANAGE A SERVICE DESK, CALL CENTER, CLINIC INTAKE, OR RETAIL FLOOR AND AN AI NOTE-TAKER IS ON YOUR ROADMAP, WRITE APPLE'S THREE CHOICES INTO YOUR REQUIREMENTS BEFORE PROCUREMENT, NOT AFTER: dual consent (worker AND customer, with a real refusal path that does not penalize either), no retention of raw audio, and no manager access to transcripts. Vendors will default to the opposite of all three because retention and manager dashboards are what they sell.
Gizmodo (2026-07-19, Mike Pearl), reporting Mark Gurman's Bloomberg newsletter: Apple has begun a limited pilot called 'Live Notes' that auto-transcribes Genius Bar conversations and generates AI summaries stored in the repair session record on employee iPads · PetaPixel (2026-07-21) on the same Gurman report: 'Both the employee and the customer must consent to the recording — otherwise, Live Notes won't be activated' · Gurman via PetaPixel: 'The company has indicated that the recordings themselves won't be saved and managers don't receive the transcripts'; employees 'can edit both the transcript and summary before saving them to Apple's internal Genius Bar system' · Gizmodo: Apple Store workers' stated concern is that 'something almost exactly like this could be used for monitoring and evaluating employees' · Scope: a small number of Apple Store locations; stated purpose is saving Genius Bar workers documentation time · primary source
4. Samsung switched on an opt-in AI Health Assistant for US users inside the Samsung Health app on July 21 — it reads your sleep, activity, nutrition, mindfulness and vitals together to explain how each affects the others, with recommendations Samsung says were validated by physicians and certified health coaches, and it is explicitly barred from diagnosing you.
The constructive part is not the chatbot, it is the joining. Most people wearing a watch or a ring already have years of health data they cannot interpret, because the numbers arrive in separate silos — sleep here, steps there, heart rate somewhere else — and the useful signal is almost always in how one affects another. Samsung's Health Assistant, live in US beta on July 21, 2026 across Samsung Health on phones, Galaxy Watches and the Galaxy Ring, is built to read those five pillars against each other and surface the pattern a person would not spot alone; Engadget's example is the assistant flagging late-night smart-light usage as a plausible cause of poor sleep, because it can also see the connected home. Samsung says the guidance 'has been validated by a team of physicians and certified health coaches.' Just as important is the line the product does not cross: it will not offer medical advice, suggest treatments, or diagnose conditions — the boundary that separates a wellness tool from an unapproved medical device, and the one the FDA-cleared systems we have covered this month (UpDoc's insulin management, iHealthScreen's diabetic-retinopathy screening) had to go through a clearance process to step over. The honest caveat belongs in the story: to do the joining, the assistant reads beyond health data — your calendar and smart-home devices included — which is a real expansion of scope, and the reason opt-in matters here rather than being a footnote.
Key fact: IF YOU OWN A GALAXY PHONE, WATCH, OR RING AND HAVE BEEN SITTING ON YEARS OF UNREAD HEALTH DATA, THE BETA IS OPT-IN AND US-ONLY RIGHT NOW — open Samsung Health and enable Health Assistant, then ask it a cross-pillar question ('why has my sleep quality dropped this month?') rather than a single-metric one, because the cross-domain correlation is the only thing it does that your existing dashboards do not.
Engadget (2026-07-21): Samsung launched 'Health Assistant' in beta on July 21, 2026 for eligible US users, integrated with Samsung Health and working with phones, Galaxy Watches and Samsung's smart ring · Engadget: the assistant monitors five wellness pillars — sleep, activity, nutrition, mindfulness and vitals — 'showing how each category influences the others' · Engadget, quoting Samsung: recommendations 'have been validated by a team of physicians and certified health coaches' · Engadget, the hard boundary: 'It won't offer medical advice or suggest treatments, and it can't diagnose health conditions' · Engadget: the assistant also draws on 'information from things like your calendar and even smart home devices' — e.g. flagging late-night smart-light usage as affecting sleep — and the tool is opt-in · primary source