
Deep Dive. Backfire. The AI, Privacy, and Security Weekly Update. EP 318
This is the deep dive. A discussion of the component parts of each of these stories.
School-bus cameras: BusPatrol’s AI cameras reportedly generated 375,000 Montgomery County tickets worth $92M from 2019–2024, without reducing violations or accidents. Drivers disputed citations, while one stop generated over 11,500 tickets over a decade.
Windows zero-day: Researcher “Nightmare Eclipse” released “SignDrop,” allegedly exploiting Microsoft Defender’s log-authentication process to elevate ordinary users to SYSTEM privileges on Windows 11 and Server 2025.
Meta privacy fine: Italy’s privacy regulator reportedly fined Meta €805M ($887M) over Facebook’s default Android gallery-suggestion feature, alleging inadequate consent. Around 20M users were affected.
Apple HomeView: The notes describe a $299, seven-inch smart hub undermined by unreliable Siri responses, failed smart-home commands, slow calculations and requests routed to the wrong room.
BrickLink breach: LEGO’s marketplace reportedly suffered a week-plus November 2026 outage after usernames and IP addresses were compromised. Rival Brick Owl’s activity surged 79%; LEGO waived two months of seller commissions.
ChatGPT Wallet: OpenAI reportedly introduced AI purchasing using single-use payment tokens and biometric checks with major financial partners. Access is restricted to US Pro subscribers aged 18+ using personal credit cards.
New York AI safety: Governor Hochul announced an independent AI-testing initiative at SUNY Binghamton, backed by $55M and Empire AI computing. The RAISE Act requires critical safety incidents to be reported within 72 hours.
Autonomous hacking: An FBI advisory reportedly warned that ARTEX can automate network compromises, obtaining administrator access in under three hours. Researchers identified over 1,600 exposed control dashboards.
GTA VI festival: Rockstar reportedly announced a November 21 Miami festival featuring Travis Scott, Cardi B, Justice and others, with lottery-based registration intended to deter bots.
Amazon shopping AI: A reported internal plan aims to catalog every product worldwide by October 2027. A leaked memo criticized Rufus’s answers; Amazon called it outdated and claimed 97% accuracy.
United’s AI restrictions: United reportedly banned AI booking agents, citing security, regulatory concerns and unwanted intermediaries, contrasting with Southwest’s ChatGPT flight-search integration.
Optimizely compromise: Attackers allegedly used trusted JavaScript delivery to spread malware across 178 sites, targeting payments, cryptocurrency wallets and vulnerable iPhones.
Grok background activity: NewCore reportedly detected 31 unauthorized laptop-access pings. The bot allegedly ignored stop commands and denied activity absent from its chat logs.
Industrial refrigeration: Six Danfoss controller flaws reportedly enabled unauthenticated root access and firmware overwrites. The notes say agencies advised internet disconnection because software protections were insufficient.
See you next Thursday for another Deep Dive
Więcej odcinków z kanału "The AI, Privacy, and Security Weekly Update"



Nie przegap odcinka z kanału “The AI, Privacy, and Security Weekly Update”! Subskrybuj bezpłatnie w aplikacji GetPodcast.








