> daily_signal(2026_07_18)

San Francisco went after Apple and Google directly over AI apps that fake nude photos, while the Senate protected a Medicare AI that's already doubling how long seniors wait for care.

PickBits Daily Signal · Saturday, July 18, 2026

By Mark Pickering · 9 min read · July 18, 2026

// tl;dr

San Francisco's city attorney sent Apple and Google cease-and-desist letters over 13 AI "nudify" apps, and both companies acted within days. Apple pulled 3 apps and is terminating those developers, with 4 more under review; Google suspended all 5 apps named. The letters landed right after the federal Take It Down Act, criminalizing exactly this conduct, cleared the Senate unanimously.
TikTok began testing an opt-in likeness-detection tool that scans for AI-generated videos wearing a creator's face without permission, following YouTube's exact template. Creators verify through Jumio (a real-time selfie plus ID check) before TikTok starts scanning. It's currently limited to a small group of US creators.
The Senate voted 46-50 along party lines to keep a Medicare AI prior-authorization pilot running in six states, with no exit date set. Sen. Maria Cantwell's own data shows approval times have more than doubled since the program started. Separate reporting adds a fact the Senate floor debate didn't mention: the private AI firms making these calls are paid based on the claims they deny.
Jennifer Doudna's lab at UC Berkeley used an AI protein-design model — the same category of AI that won the 2024 Nobel Prize in Chemistry — to invent a synthetic CRISPR enzyme that doesn't exist in nature. The AI-designed enzyme matched or beat its natural counterpart on editing efficiency while showing the lowest off-target error rate of any variant tested.

We covered a version of today's first fight back in June, when a UK lawmaker sued xAI over Grok-generated nude deepfakes of her. That case was about the AI company. This week San Francisco skipped the AI company and went straight at Apple and Google, the app stores actually hosting the tools, while TikTok raced to build its own version of a tool YouTube already shipped. Then the Senate did the opposite of protecting anyone: it voted to keep a Medicare AI's authority over seniors' care exactly as it is. The one bright spot today: Jennifer Doudna's lab used the same kind of AI that just won a Nobel Prize to build a piece of CRISPR nature never got around to.

Today: Apple and Google got sued over nudify apps, TikTok played catch-up with YouTube, the Senate sided with an AI that denies Medicare claims, and Doudna's lab built a piece of CRISPR nature missed.

1. San Francisco skipped the AI company and went straight after Apple and Google over nudify apps.

Both companies had 28 days to respond. Neither needed them.

San Francisco City Attorney David Chiu sent cease-and-desist letters to Apple and Google this week over 13 AI "nudify" apps — face-swap tools built to strip clothing from real photos without consent — 8 targeting the App Store and 5 targeting Google Play. The letters accuse both companies of "aiding and abetting" the spread of non-consensual intimate images and argue they have "likely made millions of dollars in fees" hosting the apps, citing a January Tech Transparency Project report that pegged the nudify-app economy at 705 million downloads and $117 million in revenue across both stores. Both platforms moved fast: Apple has removed 3 apps and is terminating those developers' accounts, with 4 more under review; Google confirmed all 5 Play Store apps named were suspended. The letters gave both companies 28 days to respond or face civil penalties under California law.

The timing isn't a coincidence. This same week, the federal Take It Down Act — which criminalizes publishing a non-consensual AI-generated nude image — cleared the Senate unanimously. But it's California's own law, on the books since before this week, that actually gave Chiu's office its leverage. Both companies moved within days of getting that letter. They could always have caught these apps; nobody had bothered to look until a city attorney made them.

TechCrunch July 17 2026 San Francisco City Attorney David Chiu sent cease-and-desist letters to Apple and Google over 13 AI nudify face-swap apps 8 on the App Store 5 on Google Play accuses both companies of aiding and abetting distribution of nonconsensual intimate images citing January Tech Transparency Project report of 705 million downloads and 117 million dollars in revenue Apple removed 3 apps terminating developers 4 under review Google suspended all 5 apps named 28 days to respond or face civil penalties under California law — techcrunch.com · July 17, 2026

Why this matters: If an AI-generated image of you is circulating without your consent, you have more leverage than the app store itself provides. California's law against non-consensual deepfake pornography is what forced this action, and most states now have an equivalent. Action this week: Report the specific app to Apple or Google Play citing non-consensual intimate imagery by its legal name, not as generic "inappropriate content." A legally specific complaint moves faster. If you run an app store or any marketplace with third-party AI tools, don't wait for the next city attorney's letter. Start looking for this yourself.

techcrunch.com: Apple and Google ordered to purge nudify apps from app stores (July 17, 2026)
wired.com: San Francisco demands Apple and Google delete AI nudify apps from app stores (July 2026)
tech.yahoo.com: San Francisco demands Google and Apple remove nudify apps from their stores (July 2026)

2. TikTok is testing the exact tool YouTube already built to catch deepfakes of your face.

Right now it only covers a small group of creators.

TikTok began testing an opt-in "likeness detection" tool this week that scans for AI-generated videos using a creator's face without permission. Creators who want protection first verify their identity through Jumio — a real-time selfie plus an ID check; TikTok says it doesn't retain the ID documents, only the facial data used to match against AI content on the platform. Once verified, TikTok scans for matches and lets the creator decide whether to report a post as impersonation. TikTok US spokesperson Zachary Kizer confirmed the test is limited to a small group of US creators — a pilot, not a platform-wide rollout.

TikTok isn't inventing this category. It's running YouTube's playbook almost exactly: YouTube built its own likeness-detection tool after months of quiet testing, then expanded it to eligible creators over 18 and extended similar protection to celebrities and talent agencies. The one thing TikTok's version adds that YouTube's doesn't is that upfront Jumio identity check. Now the two largest short-video platforms both have a likeness-detection tool. This is just standard now. Nobody's going to advertise it as a feature. Neither platform protects you yet if you're not a paid creator. An ordinary person whose face shows up in an AI video still has nothing to point to but a general abuse report.

Why this matters: If you're a US creator worried about your likeness being used in AI-generated videos, small pilots like this tend to expand by invite before they expand by open enrollment, and being verified early means TikTok is scanning on your behalf sooner. Action this week: Check your eligibility for the pilot now, and read TikTok's specific retention claim (ID documents not kept, facial data used only for matching) before opting in, since that's the sentence to hold the company to if the policy changes. If you're not a verified creator, push TikTok and YouTube through their public feedback channels to extend likeness verification beyond monetized accounts. The technical capability clearly already exists.

digitaltrends.com: After YouTube, TikTok is testing its own AI likeness detection tool (July 2026)
theverge.com: TikTok is testing an AI likeness detection tool (July 17, 2026)

3. The Senate had the chance to end a Medicare AI that's already doubling how long seniors wait for care. It didn't.

One senator's own data is the number to remember here.

Back in May, we tracked CMS's growing embrace of AI in Medicare when it started paying for AI agents to help manage chronic conditions between doctor visits. This week that same agency's AI got to decide who gets denied care. On Thursday, the Senate voted 46-50, exactly along party lines, on a resolution to end the WISeR ("Wasteful and Inappropriate Service Reduction") model — a CMS pilot using AI to approve or deny prior authorization for Medicare patients in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. Fifty Republicans blocked the resolution; 46 Democrats voted to end the program. It keeps running with no exit date. Sen. Maria Cantwell's own analysis of Washington state data found patients waiting 2 to 4 times longer for procedures their doctors recommended since WISeR started, from roughly two weeks to four to eight weeks. CMS Administrator Dr. Mehmet Oz defended the program as "committed to crushing fraud, waste and abuse."

Nobody said this part on the Senate floor: separate local reporting found Medicare is contracting private AI firms to make these calls, and those firms get paid based on how many claims they deny. Both parties just voted with that arrangement still in place. There's still a second track that could change this: the House Appropriations Committee already voted unanimously in June to block WISeR's funding in next year's spending bill, but that still needs the full House and Senate to pass. Until then, the program keeps running exactly as it has since January.

STAT News July 16 2026 Senate voted 46-50 along party lines Thursday blocking Congressional Review Act resolution to end WISeR Medicare AI prior authorization pilot Senator Maria Cantwell data shows approval times more than doubled from 2 weeks to 4 to 8 weeks CMS Administrator Mehmet Oz defends program as fighting fraud waste and abuse House Appropriations Committee voted unanimously in June to block WISeR funding still needs full House and Senate passage program continues in six states Arizona New Jersey Ohio Oklahoma Texas Washington — statnews.com · July 16, 2026

Why this matters: If you or a family member are on traditional Medicare in Arizona, New Jersey, Ohio, Oklahoma, Texas, or Washington, and your doctor recommends a covered service, this 4-to-8-week wait isn't going away. It's just how it is now. Action this week: Ask your provider's office to file for prior authorization the moment a recommendation is made, not when the procedure is scheduled. If you're denied, appeal, and ask explicitly whether an algorithm or a person made the initial call, since that changes what your appeal should target. Track the House Appropriations rider; if it clears both chambers, WISeR funding could end before its six-year term does.

statnews.com: Senate blocks resolution to end Medicare AI prior-authorization pilot (July 16, 2026)
newsweek.com: Medicare update, list of senators voting against repeal of major AI change (July 2026)
healthcaredive.com: House Appropriations Committee votes to block AI prior-authorization pilot WISeR (June 2026)

4. Jennifer Doudna just pointed Nobel-winning AI at her own invention, and built something nature never did.

Finally, some good news.

In 2024, the Nobel committee gave the Chemistry prize to the scientists — including David Baker — who taught AI to design entirely new proteins from scratch. This week, Jennifer Doudna, who won her own Nobel for co-discovering CRISPR-Cas9, pointed that same category of AI at her own field. Her team at the Innovative Genomics Institute and UC Berkeley used an inverse protein-folding model — AI that starts from a target 3D structure and reverse-engineers the amino-acid sequence that would build it — constrained by evolutionary data from the natural TnpB family of CRISPR-Cas12-like proteins, to design synthetic nucleases they call SynTnpBs. Cryo-electron microscopy of the best-performing variant showed the AI had introduced new electrostatic and hydrogen bonds at the guide-RNA-to-DNA interface, contacts a natural evolutionary search never found. In gene-editing assays across animal and plant cells, the top SynTnpB variant matched or exceeded its natural counterpart's editing efficiency while showing the lowest off-target activity of any variant tested, published in Science this week.

Every CRISPR therapy has to solve two problems at once: cut where you mean to, and don't cut anywhere else. Off-target editing specifically has been the main safety brake keeping CRISPR out of approved human therapies, and this enzyme improved both at the same time. Doudna's team can now point this same AI method at the next gene-editing bottleneck, not just this one enzyme. No specific disease application or clinical timeline was named yet; this is foundational tool-design research, not a therapy announcement.

The Scientist July 16 2026 Jennifer Doudna Innovative Genomics Institute UC Berkeley used AI inverse protein-folding model constrained by TnpB family evolutionary data to design synthetic CRISPR nucleases called SynTnpBs cryo-electron microscopy shows new electrostatic and hydrogen bonds at guide RNA to DNA interface top variant matched or exceeded natural enzyme editing efficiency with lowest off-target activity published in Science establishes general framework for AI-designed non-natural nucleases — the-scientist.com · July 16, 2026

Why this matters: If you work in gene-editing research or therapeutic development, the reusable part of this result is the method, not just the enzyme: an inverse-protein-folding-plus-evolutionary-constraint approach that beats its natural counterpart on both efficiency and off-target rate is a template applicable beyond the TnpB family. Action this week: If off-target editing has been the problem holding your work back, this is worth testing directly. If you follow AI-for-science as a category, watch for the follow-on papers. If this is really a general framework and not a one-off, the same method should work on a second, unrelated nuclease family within the year. That's what would prove it.

the-scientist.com: A synthetic nuclease boosts CRISPR editing efficiency (July 16, 2026)
nature.com: AI designs synthetic CRISPR enzymes that outperform nature's own (July 16, 2026)

» What to watch this week

Whether Apple and Google's app-review sweeps expand beyond the 13 named apps before the 28-day deadline. The letters targeted a specific list; whether either company actually fixed this, or just responded to one letter, shows up in whether new nudify apps keep slipping through.
Whether TikTok's likeness-detection pilot expands past its current small group of US creators, and on what timeline. YouTube's own rollout took months from quiet testing to general availability. Watch whether TikTok moves faster now that the template already exists.
Whether the House Appropriations rider blocking WISeR funding actually clears the full House and Senate. A unanimous committee vote in June is not the same as a floor vote; this is the track that could still end the program the Senate just chose not to.
Whether a second AI-designed nuclease family follows the SynTnpB result. Doudna's team frames this as a general framework, not a one-off. The next twelve months of follow-on papers either confirm that or narrow the result back down to one enzyme.

Tomorrow's signal lands here.