Weekly Tech & AI Roundup: Meta's $25B Privacy Fine, FCC's Huawei Camera Ban & Claude Duet (Oct 19–25, 2026)

The biggest number of the week: $25 billion, the civil penalty a New Mexico judge ordered Meta to pay for misleading Facebook users about their data — the largest state verdict against a company in U.S. history. Meanwhile, regulators drew a new line around surveillance hardware, a budget camera brand had its biggest launch week of the year, and Anthropic rolled out an AI agent that works while you're away from your desk.

The week in 60 seconds

• A New Mexico judge ordered Meta to pay $25 billion over Facebook privacy violations on October 21, 2026 — the largest state verdict against a U.S. company on record.
• The FCC blocked its first two devices under a new chip-level rule on October 22, 2026, both carrying Huawei-linked HiSilicon surveillance chips with face-detection features.
• Wyze launched a $69.98 solar-powered camera (Oct. 20) and a $99.98 pan-tilt bulb camera (Oct. 22) during its "Wyze Week" sale.
• Anthropic opened a waitlist for Claude Duet on October 21, 2026, a proactive AI agent that keeps working on tasks after you close the app.
• A single hacker used off-the-shelf AI agents to breach 100+ companies for about $5,000 total, Reuters reported October 19, 2026.
• Microsoft declared that enterprise AI agents should be treated as "insider" risks, rolling out new monitoring controls on October 19, 2026.

Smart Home & Security Cameras

Wyze ran its annual "Wyze Week" promotion from October 19–23, 2026, debuting two new outdoor cameras. On October 20, the company launched the Wyze Solar Cam, a $69.98 wire-free 2K camera that the company says stays charged on as little as one hour of direct sunlight a day. Two days later, Wyze followed with the Wyze Bulb Cam Pan Color, a $99.98 2.5K pan-and-tilt camera built into a screw-in light-socket smart bulb. For your home: both launches reinforce how cheap dedicated camera hardware has gotten — but if you already own a spare phone, apps like iCameraPlus let you get continuous recording and a private cloud archive without buying any new hardware at all.

On October 22, 2026, the FCC's Office of Engineering and Technology announced it had blocked a dashboard camera and a wireless video transmitter from the U.S. market — the first enforcement of a new rule, effective October 13, 2026, that bans devices containing chips from companies on the FCC's Covered List, regardless of the device's brand. Both products used HiSilicon chips, a Huawei subsidiary, which the manufacturer advertised as "designed for the surveillance market" with built-in face-detection algorithms, according to the FCC's own order. The new rule closes a loophole that let restricted components ship inside rebranded, non-Chinese-labeled products. For your home: this doesn't touch mainstream U.S.-sold camera brands, but if you bought a bargain dashcam or wireless camera transmitter from an unfamiliar brand online, it's worth checking whether it's since been flagged — our guide on checking your camera against known vulnerability lists covers the steps.

Privacy & Surveillance

Meta must pay New Mexico $25 billion — the largest state verdict against a company in U.S. history. On October 21, 2026, a judge issued a civil penalty order against Meta after a jury found more than 43 million violations of New Mexico's Unfair Practices Act tied to the Cambridge Analytica scandal, according to the New Mexico Department of Justice. The money is earmarked for the state's education fund, and Meta has said it will appeal.

"One of the largest, if not the single largest, state verdict against a company in American history." — New Mexico Attorney General Raúl Torrez, via the New Mexico Department of Justice

For your home: the case isn't about cameras, but it's a reminder that regulators are now willing to impose fines large enough to matter to a company's stock price when it misleads users about who controls their personal data — a question worth asking of any company storing your footage in the cloud, discussed further in our privacy practices overview.

A hacker breached over 100 companies using AI agents for about $50 a target. A Reuters investigation published October 19, 2026 found that a Chinese-speaking operator pointed autonomous AI agents — built on the open-source Cairn platform and running a Claude Code harness driving DeepSeek's V4.1 Flash model — at more than 1,200 e-commerce sites starting September 25, breaching over 100 organizations including Deere, SharkNinja and Cummins for roughly $5,000 in total token spend, as reported by Reuters. The attacker's instructions reportedly fit in fewer than 100 characters; the agents handled reconnaissance and exploitation with little further direction. For your home: this is aimed at corporate web servers, not home devices, but it illustrates how cheap automated attacks have become — all the more reason to keep your router firmware and any camera or smart-home app updated rather than running on defaults.

Artificial Intelligence

Anthropic opened a waitlist for Claude Duet on October 21, 2026, a research preview of a proactive AI agent built on Claude Opus 5.5 that prepares daily briefings and keeps working on connected apps — Gmail, Slack and more than 6,000 other integrations — even after a user closes the app, as CNBC reported. The preview is initially limited to Pro and Max subscribers in the U.S., and Anthropic says the agent requires explicit approval before sending messages, making payments or deleting data. The launch follows Meta's Muse and OpenAI's Dots, both introduced in September 2026, as major AI labs race to build assistants that act without constant prompting.

Separately, Microsoft said on October 19, 2026 that it now treats enterprise AI agents as potential "insider" security risks, rolling out Guardian Agents and dynamic permission controls in a post on its official security blog. The company expanded its agent-discovery tooling to cover 60 types of local coding agents, including Claude Code, OpenAI Codex and GitHub Copilot, and argued that "frontier models can't be trusted" to self-police their own actions inside company systems, per TechCrunch's coverage of the announcement.

Notable Releases

Meta reported third-quarter 2026 earnings on October 21, raising its full-year capital-expenditure outlook to between $135 billion and $145 billion to fund its Muse AI buildout, and announced a new nonvoting Class C stock structure, according to the company's earnings call transcript. Quarterly capex hit $34.9 billion, up from $19.4 billion a year earlier, underscoring how much of this week's AI news — from Claude Duet to Microsoft's agent-security push — is being driven by a continuing surge in infrastructure spending across the industry.

Sources

• New Mexico Department of Justice: $25 Billion Penalty Against Facebook
• FCC: Order Blocking Two Devices Containing Foreign Surveillance Chips
• Tom's Hardware: FCC Bans First Two Devices Under New Chip-Level Rule
• Business Wire: Wyze Unveils $69.98 Solar Cam
• Business Wire via Morningstar: Wyze Bulb Cam Pan Color
• Reuters via U.S. News: How One Hacker Used AI Agents to Break Into Scores of Sites
• Anthropic: Introducing Claude Duet
• CNBC: Anthropic Announces Preview of Claude Duet
• Microsoft Security Blog: Agents Are Your Newest Insiders
• TechCrunch: Microsoft Says AI Agents Are Like Insider Threats
• The Motley Fool: Meta Q3 2026 Earnings Call Transcript