Weekly Briefing: Privacy

The clearest lesson from the week was that the central privacy risk in police surveillance no longer lies only in the act of collection. It lies in everything that follows. The Los Angeles Police Department declined to renew its Flock contract after an inspector general audit raised concerns about vehicle misidentifications and unresolved rules for ownership, security and sharing. In Milwaukee and Greer, South Carolina, documented misuse led to a guilty plea, firings and tighter access controls. Taken together, these cases show surveillance governance moving from abstract concern to an observable record of searches, errors and employee conduct.

That shift helps explain why procurement may become a more immediate restraint than sweeping legislation. As The Record reported, LAPD wanted enforceable terms before continuing with Flock, including clearer control over how vehicle data is handled and shared. Yet Cleveland considered an extension, Kenton County weighed six new readers, and a Louisiana parish deployed 50 cameras. The technology is not simply advancing or retreating. Instead, adoption is becoming conditional on whether agencies and vendors can defend the rules surrounding use.

More revealing than the number of cameras was the growing importance of auditability. Milwaukee did not discover an officer’s 179 searches internally; a person who was tracked found evidence through a public transparency site. Greer, by contrast, used Flock’s audit assistance to investigate suspected policy violations and fired two officers. WIRED’s report that San Francisco police drone feeds had been publicly accessible added a different version of the same problem: a surveillance system may be authorized for public safety and still create privacy exposure through weak access controls. The common vulnerability is the operational layer connecting collection, storage and human access.

AI privacy followed a similar pattern, although the evidence was more fragmented. Meta withdrew a feature that let users generate images involving tagged public Instagram accounts after criticism of its default-on design. Separately, WIRED reported that face-recognition code associated with a disputed smart-glasses capability had appeared in, and then been removed from, the Meta AI app. Google introduced account-linked avatars based on a user’s face and voice. These developments do not establish a single regulatory direction, but they do make one point clearer: consent choices made during product design are becoming consequential before courts or regulators have settled how older biometric and likeness rules apply.

The week also exposed a harder conflict between safety and data minimization. Meta began offering opt-in parental alerts when supervised teenagers discuss self-harm or suicide with its AI, while analysis from the Future of Privacy Forum described state laws requiring some chatbots to detect suicidal ideation and direct users toward crisis resources. Those duties may be well-intentioned, but they require systems to infer and sometimes retain exceptionally sensitive mental-health information. AI safety rules are therefore beginning to create privacy obligations of their own, rather than merely limiting harmful outputs.

Breach accountability supplied a more established form of pressure. A multistate settlement over the 2023 23andMe breach focused attention on missing defenses against credential stuffing, weak monitoring and the way access to a small number of accounts expanded through genetic-relative features. The same week brought phishing aimed at LastPass and Bitwarden users, malicious npm packages targeting developer secrets, and disclosures involving healthcare and vendor-held customer data. Not every credential incident was a corporate breach, but the combined lesson was consistent: identity controls and limits on secondary access increasingly determine the scale of privacy harm.

Biometric infrastructure continued to expand without producing a comparably coherent accountability model. The UK and France added border capacity after biometric kiosks malfunctioned, while Gibraltar’s new arrangements paired easier physical movement with Schengen-style biometric processing and expanded facial-recognition coverage. These cases show that operational friction can slow biometric systems without reversing their underlying policy momentum. For now, borders, policing and consumer devices remain separate legal arenas even as they converge technically around faces, movement and identity.