Daily Surveillance Briefing: November 13, 2026

A UK police watchdog found that forces cannot show retrospective facial recognition makes their investigations better, that officers acted on matches without corroborating them, and that deleted custody images stayed searchable. A federal appeals court ordered the Justice Department to narrow its LinkedIn dragnet for journalists. New York City passed a surveillance pricing ban. Fourteen governments adopted the first global guidelines for commercial spyware vendors. Amazon's Photos biometric case was certified as a class action. Google's Gemini smart glasses leave video unencrypted on its servers. A breach at Fireworks AI took Notion AI chat transcripts with it.

State of Surveillance

Today in Surveillance:

  • A UK police watchdog says forces cannot show retrospective facial recognition helps, and that officers acted on matches without checking them. UK law enforcement still runs more than 300,000 searches a year [1][2].
  • A federal appeals court ordered the Justice Department to narrow its LinkedIn dragnet for journalists. Judge Harvie Wilkinson III wrote that many of the users swept up "have no conceivable relationship" to the leak probe [3].
  • New York City passed a surveillance pricing ban. The city could fine companies up to $1,000 for each transaction that violates it [4][5].
  • Fourteen governments adopted the first global guidelines for commercial spyware vendors. The rules are voluntary, carry no penalties, and were endorsed by fewer than half the 43 countries there [6][7].
  • A federal judge certified the Amazon Photos biometric class action, after Amazon identified 339,847 Illinois-linked accounts with image recognition enabled [8].
  • Google's Gemini smart glasses ship December 10 with face redaction, no live facial recognition, and video left unencrypted on Google's servers. [9]
  • A breach at Fireworks AI took Notion AI chat transcripts with it. Notion says about 1% of its customers were affected [10].

Also today: Trezor confirmed that three Safe 5 wallets bought from the reseller CryptoBilis carry an implant built to steal recovery phrases, and Trump nominated his pardons adviser Gary Lawkowski as ODNI general counsel [11][12].

A Watchdog Found Police Treating Face Recognition Matches as Evidence

His Majesty's Chief Inspector of Constabulary and Fire & Rescue Services published a 58-page inspection of retrospective facial recognition this week, and the finding is blunt: UK police cannot show the technology makes their investigations better. Commissioned by Home Secretary Shabana Mahmood in November 2025, the report says forces "couldn't clearly show performance improvements from using RFR". Despite that, UK law enforcement makes more than 300,000 retrospective search requests a year [1].

The risk is what happens to a match after it lands. The report says: "Some investigators didn't know that RFR results were intelligence rather than evidence and required further enquiries before officers took enforcement action." Inspectors found investigators did not always corroborate potential matches before acting, increasing the risk of wrongful identification and arrest. Roughly one-third of officers surveyed were either wrong or unsure whether a match meets the legal requirement for evidence of identification. It does not. Nearly two-thirds of the 779 Police National Database users surveyed by the Home Office could not recall any facial recognition training in the past year [1][2].

Custody images that should have been deleted were still available for searches, and the National Physical Laboratory found in 2024 that the database algorithm returned more false positive matches for certain demographic groups at the threshold in use; the threshold was raised, then the National Police Chief's Council told chief constables about the change in November 2024 and lowered the setting again later that same month. HMICFRS says it "couldn't establish with certainty whether the bias identified in the algorithm had contributed to any wrongful arrests" [1].

Big Brother Watch's Jasleen Chaggar said the report shows how officers use the output: "The revelations that officers are treating facial recognition as bona fide evidence rather than a probability score is alarming".[2] Our coverage of the [Met's expansion](/news/uk-met-police-facial-recognition-expansion-2026), the [regulators' patchwork verdict](/news/uk-biometrics-commissioners-facial-recognition-patchwork-ineffective-2026) tracks how the policy was built. [Angela Lipps spent five months in jail](/news/angela-lipps-facial-recognition-wrongful-arrest-five-months-jail-2026) on a facial recognition error.

An Appeals Court Told the DOJ to Stop Sweeping Up Random Journalists

A unanimous three-judge federal appeals panel ordered a district court to narrow the administration's request for 18 months of LinkedIn communications from journalists. Writing for the panel, Judge J. Harvie Wilkinson III said: "Many of the users identified have no conceivable relationship to the leak investigation." Instead, the users were "affiliated with media outlets, universities, and law firms", not the leak targets [3].

The order was not about whether the leak probe is legitimate. It was about the paperwork. "It is not appropriate for district courts to approve requests purely because the records requested are logistically feasible to compile," Wilkinson wrote. The appeals court also took aim at the blanket year-long secrecy order that left journalists unaware their communications were under review, and told the government to justify it rather than rely on the blanket [3]. Our reporting on [Google handing a student journalist's bank numbers to ICE without a judge](/news/google-ice-subpoena-student-journalist-bank-data-2026) covers the other end of the same chain.

New York City Passed a Surveillance Pricing Ban

The New York City Council voted on Thursday to ban surveillance pricing, the practice of charging people different prices for the same goods based on what algorithms know about them. Speaker Julie Menin's Introduction 891-A would prohibit any company from setting prices, in whole or in part, by algorithm that uses personal data to offer different customers different prices for the same goods. Gothamist's Elizabeth Kim reports the city could fine companies up to $1,000 for each transaction that violates it [4][5].

The mechanism is location tracking. In one investigation cited by the Council, the price of a television at one retailer went from $499.99 to $599.99 when the consumer pulled into the store's parking lot [5]. Menin said: "It's a practice that worsens the affordability crisis, and today it's a practice that we are going to be ending."[4] Mayor Zohran Mamdani backs the ban, calling surveillance pricing "just another way corporations are finding to squeeze working people" [4]. Our [state-by-state tracker](/news/surveillance-pricing-24-states-ban-algorithmic-pricing-tracker-2026) has the wider map, and [New York's state ban drew a lawsuit](/news/surveillance-pricing-new-york-one-fair-price-act-signed-washington-post-class-action-2026) of its own.

Fourteen Governments Wrote the First Global Rules for Spyware Vendors

The UK and France led fourteen governments in endorsing the first set of industry guidelines for commercial cyber intrusion capabilities, the companies that build, sell and finance the tools implicated in surveillance campaigns against journalists, political officials, activists and businesses. The guidelines were launched at the third Pall Mall Process conference in Paris on November 11, and the French Foreign Ministry called them "the first global standard of responsible conduct" [6][7].

The ask is modest: vet customers, build safeguards against misuse, halt sales when serious misuse becomes a credible risk, and "consider" remedies for victims of human rights abuses. There are no penalties. The numbers undercut the claim of global consensus: the conference drew representatives from 43 countries and fewer than half endorsed the guidelines. Twenty-seven countries have signed last year's code of practice, and it is not clear why the remaining 15 did not endorse the new guidelines [6]. Our coverage of [Greece's Predator convictions](/news/greece-predatorgate-intellexa-spyware-convictions-2026) tracks the gap between rules and prosecutions.

Amazon Photos Biometric Suit Certified as a Class Action

A federal judge certified a class action against Amazon over the face-recognition feature in Amazon Photos. The case, Hogan v. Amazon.com, Inc., covers Illinois residents whose faces were processed in photos uploaded to their own accounts between October 18, 2016 and January 31, 2023. Amazon identified 339,847 Illinois-linked accounts with image recognition enabled during that period, though not all necessarily qualify for the class [8].

Amazon pointed to its Terms of Service, but the court found the terms never used the words "biometric identifier", "biometric information" or "facial recognition", and that users could turn on image recognition without ever seeing the required notices. Illinois' Biometric Information Privacy Act sets statutory damages at $1,000 for negligent violations and $5,000 for intentional or reckless ones, and Amazon is already fighting BIPA suits over Just Walk Out retail biometrics and Alexa voice data.[8] [Ring's Familiar Faces feature scans your neighbours](/news/amazon-ring-familiar-faces-class-action-facial-recognition-consent-2026) too.

Google's Smart Glasses Redact Faces. They Also Leave Video Unencrypted.

Google put its first Gemini smart glasses on preorder on Thursday, built with Samsung and sold through Warby Parker frames starting at $395 and Gentle Monster styles starting at $449. The privacy pitch is the selling point: a recording light that stays on whenever the camera is live, a refusal to record if that light is covered or tampered with, a physical off switch, no Play Store app allowed to livestream video from the glasses, no real-time facial recognition, and automatic face redaction before video reaches Google [9].

Ars Technica's Ryan Whitwam found the gap between the pitch and the engineering. "Data on the glasses is encrypted, and it remains that way during transit to the cloud. However, it's not encrypted on Google's servers," he writes. Private AI Compute, which isolates data during processing so not even Google can access it, is not on the glasses yet. The devices ship December 10 [9]. [Meta's glasses, which stripped facial recognition out of the app](/news/meta-strips-facial-recognition-smart-glasses-app-2026), are the comparison that matters.

A Breach at Fireworks AI Took Notion's Chat Transcripts With It

Notion AI customers who used the Insights feature had their chat transcripts taken by attackers who compromised Fireworks AI's cloud provider and reached a development system, the model platform Notion used. TechCrunch's Zack Whittaker reports that the attackers also took customer billing information and datasets. Notion says the breach affected around 1% of its customers, and that the stolen transcripts include "any input to or output from the model that was written to the chat" [10].

Fireworks believes the hackers are financially motivated, selling model access to unauthorized third parties. Notion disconnected from Fireworks on November 3 and said much of the stolen data was unattributable [10]. The supply chain is the story here: a third-party provider held a copy of what people typed into a notes product, and a breach at the provider became a breach of that product's users. Fireworks says this was its second security incident in two months [10].

What to Watch

The statutory code. HMICFRS wants the Home Office to consider new legislation and a statutory code of practice governing retrospective facial recognition, and asks for a national facial recognition system by December 2027 with mandatory training by the same deadline. It also wants independent oversight of all police use of facial recognition, fresh data protection impact assessments by June 30, 2027, and local policies at every force by midway through 2027 [1].

New York's mayor. Mamdani backs the ban. Business groups warn the ban could hinder discounts and loyalty programs, and Uber's Josh Gold said it would hurt small restaurants with limited marketing budgets [4].

The spyware non-signatories, and the Notion fallout. Twenty-seven countries have signed last year's code of practice for states, and the remaining 15 did not endorse this week's industry guidelines [6]. Notion says around 1% of its customers were affected and much of the stolen data was unattributable [10].

Sources

  1. Biometric Update, Chris Burt - UK police don't know how much retrospective facial recognition helps: inspection (November 12, 2026). https://www.biometricupdate.com/202611/uk-police-dont-know-how-much-retrospective-facial-recognition-helps-inspection
  2. The Guardian, Vikram Dodd - Facial recognition errors put ethnic minorities at risk of wrongful arrest, police watchdog warns (November 12, 2026). https://www.theguardian.com/technology/2026/nov/12/facial-recognition-errors-put-ethnic-minorities-at-risk-of-wrongful-arrest-police-watchdog-warns
  3. Ars Technica, Ashley Belanger - Trump admin loses fight to spy on random journalists' LinkedIn DMs (November 11, 2026). https://arstechnica.com/tech-policy/2026/11/trump-admin-loses-fight-to-spy-on-random-journalists-linkedin-dms/
  4. Gothamist, Elizabeth Kim - NYC Council passes ban on 'surveillance pricing' that relies on personal data (November 12, 2026). https://gothamist.com/news/nyc-council-passes-ban-on-surveillance-pricing-that-relies-on-personal-data
  5. New York City Council - New York City Council Approves Landmark Legislation Banning Surveillance Pricing (November 12, 2026). https://council.nyc.gov/press/2026/11/12/3339/
  6. The Record, Alexander Martin - Britain, France set out first guidelines for commercial hacking firms (November 11, 2026). https://therecord.media/britain-france-set-out-guidelines-for-commercial-hacking-firms
  7. GOV.UK, Foreign, Commonwealth & Development Office - The Pall Mall Process Guidelines for Industry (November 11, 2026). https://www.gov.uk/government/publications/the-pall-mall-process-guidelines-for-industry
  8. Biometric Update, Chris Burt - Amazon Photos biometric privacy lawsuit certified as class action (November 12, 2026). https://www.biometricupdate.com/202611/amazon-photos-biometric-privacy-lawsuit-certified-as-class-action
  9. Ars Technica, Ryan Whitwam - Google pushes privacy angle as Gemini-powered smartglasses debut starting at $395 (November 12, 2026). https://arstechnica.com/google/2026/11/google-pushes-privacy-angle-as-gemini-powered-smartglasses-debut-starting-at-395/
  10. TechCrunch, Zack Whittaker - Cyberattack on Fireworks AI allowed hackers to access Notion customers' chat transcripts (November 12, 2026). https://techcrunch.com/2026/11/12/cyberattack-on-fireworks-ai-allowed-hackers-to-access-notion-customers-chat-transcripts/
  11. Trezor - Our CryptoBilis investigation: what we found and what we've changed (November 2026). https://trezor.io/blog/news/our-cryptobilis-investigation-what-we-found-and-what-weve-changed
  12. The Intercept, Matt Sledge - Trump Taps Campaign Lawyer and Pardon Adviser for Top U.S. Intel Agency (November 12, 2026). https://theintercept.com/2026/11/12/trump-lawyer-lawkowski-pardons-odni/