Anthropic makes first move into physical AI

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

🦾 Anthropic makes first move into physical AI

🕶️ Meta adds kill switch to smart glasses

🛡️ OpenAI and 100+ firms warn AI cyberattacks are imminent

📌 Judge blocks Pentagon's Anthropic ban

🚫 Nvidia paused its revenue-sharing deals

📺 Apple raises streaming prices for fourth straight year

Plus: 🎁 14 other news you might like, 🧰 6 tools, and 📚 5 papers.

🦾 Anthropic makes first move into physical AI LINK

Anthropic has stepped into physical AI with the Model Hardware Standard, a specification now in research preview that tells an AI how a machine works and sets safe limits, like capping a robot arm's speed and range of motion.

Vendors write an MHS file so a model can run equipment without a paper manual or expert help, and Anthropic says a Genentech scientist sent Claude a PDF experiment that the assistant carried out on its own using MHS-equipped hardware.

Amazon Web Services, Danaher, Hugging Face and Raspberry Pi are testing the standard, with a waitlist now open, though Europe's Machinery Regulation starting January 20, 2027 could treat such a file as a regulated safety component.

🕶️ Meta adds kill switch to smart glasses LINK

Meta is trying to fix the "pervert glasses" reputation of its camera-equipped Ray-Ban AI Glasses, with a new patent describing a "mute or lock" switch that can shut off the device's sensors.

The patent, filed in February and granted last week, covers "secure hardware-enforced sensor privacy systems" for augmented reality devices, with drawings that closely match Meta's current glasses, suggesting a response to its recent privacy scandals.

The move follows widespread misuse by creeps, pranksters, and bullies, plus a report that Nairobi-based contractors reviewed intimate footage; owners can already defeat the recording LED with a simple sticker.

🛡️ OpenAI and 100+ firms warn AI cyberattacks are imminent LINK

More than 100 companies, including OpenAI, Anthropic, Google, Microsoft and AWS, signed an open letter warning that AI-powered cyberattacks will soon become far more common and harder to stop as models grow more capable.

The letter, titled "A call for collective action on cyber defense," pushes organizations, governments and AI firms to treat security as a top priority and admit that current defenses won't hold against these coming threats.

It asks frontier AI companies to build monitoring and security tools, make AI agents traceable and accountable, and share methods for spotting attacks, though the warning comes after chatbots were already used to hack governments.

📌 Judge blocks Pentagon's Anthropic ban LINK

A federal judge overturned the Pentagon's decision to blacklist Anthropic as a supply-chain risk, ruling that the government violated the AI company's constitutional rights in their fight over military use of its models.

U.S. District Judge Rita Lin wrote in her 59-page ruling that "national security is not a blank check" for retaliation, noting the Pentagon kept working with Anthropic even after labeling it a risk.

The clash began this year when Anthropic refused to let the military use Claude for mass surveillance of Americans or fully autonomous weapons; the government is expected to appeal, and a separate designation is being fought in the D.C. Circuit.

🚫 Nvidia paused its revenue-sharing deals LINK

Nvidia has paused its program that gave cloud providers credit support and revenue sharing to build AI data centers, worried that the deals could draw antitrust scrutiny, The Wall Street Journal reported.

Internal talks at Nvidia questioned how far the chip designer could go in backing its customers and shaping their operations, so it froze the effort last week, though the program was not canceled and could return later.

The July initiative included partners like Sharon AI, deploying 40,000 Blackwell GPUs, and Firmus, planning up to 170,000 GPUs in Indonesia; Nvidia's earnings also revealed $108 billion in guarantees and credit deals for AI buildouts.

📺 Apple raises streaming prices for fourth straight year LINK

Apple has bumped up the price of Apple TV for the fourth year running, adding $2 to make the monthly plan $14.99, while the yearly option climbs $20 to $119.99, starting today.

The Apple One individual bundle, which packages Apple TV with Apple Music, Apple Arcade, and cloud storage, is also going up $2 to $21.95 a month, with existing subscribers given one month notice.

The hikes follow Apple loading up on content like the returning Ted Lasso and newly acquired U.S. Formula 1 rights, mirroring recent price jumps from Peacock, ESPN, Netflix, Prime Video, and YouTube Premium.

Other news you might like

• Stripe-led group is reportedly abandoning its PayPal acquisitionLINK

• Gemini Omni can now extend videos to 40 seconds and upscale them to 4KLINK

• AI’s memory crunch is coming for Android appsLINK

• YouTube now lets creators tag Amazon products and earn commissions from purchasesLINK

• Nvidia to start employee-funded US political action committeeLINK

• AI benchmarks have a trust problem and Google wants to fix itLINK

• Bluesky has a new setting to help you avoid accidentally going viralLINK

• Trump administration weighs expanding chip tariffs to laptops, consoles, and servers, report claims — January's data center exemptions may be scrappedLINK

• Google agrees to settle UK app developers' lawsuit for $353 millionLINK

• a16z creates a $1.1B ‘Machine Age’ fund to ‘accelerate the physical buildout of AI’LINK

• Microsoft cracks down on employee AI use after one worker spent $28,000 in 28 daysLINK

• US Federal Trade Commission probes YouTube over social media policiesLINK

• Always-on and self-starting AI agents might be OpenAI's next big playLINK

• Anthropic Is Quietly Cutting Google Out of the EquationLINK

🧰 Trending tools

Treebar: monitors Git worktrees and Codex agents from your MacBook notch, showing active branches and per-agent diffs with read-only local access.LINK

Open Analytics: self-hosted, privacy-friendly web analytics on Supabase, letting you query raw events with SQL, skip cookie banners, and honor Do Not Track signals.LINK

ANCBuddy for Bose QC Ultra: a macOS menu bar utility for QuietComfort Ultra headphones that toggles ANC modes and shows battery without opening the phone app.LINK

PaymentKit: manages subscriptions and routes payments across multiple processors, unifying billing metrics with no-code setup or clean APIs for fast integration.LINK

Cortex: a model-agnostic memory layer for AI agents that maintains persistent shared context by rejecting redundant writes, scoring trust, and flagging contradictions.LINK

Trama: watches your repeated macOS actions and converts plain-English descriptions into automations, no coding or workflow diagrams required.LINK

📚 Trending papers & reports

Business AI agents can be split into a freely evolving personality layer and a separate audited work layer, so companies in regulated industries keep every consequential action traceable while letting the assistant's tone drift.LINK

Real-time robot navigation lets a robot find objects it has never seen in unfamiliar spaces while the clock keeps running, boosting success rates up to 11% by treating the AI's thinking time as a real cost.LINK

Prediction confidence checks can flag which forecasting models will break when the world shifts, by spotting models that lean too heavily on one input, before you ever see live data.LINK

Chip design copilots are best understood as coordinators that connect tools across the whole design flow, not one-shot code generators that produce plausible but physically incorrect hardware and fail on industrial-scale chips.LINK

Auto-generated software testing replaces bloated test suites with a compact 32-check set that catches 99% of hidden bugs while running just 1.2 to 2% of exhaustive tests, cutting validation cost for machine-written programs.LINK

See you tomorrow for a new dose of ☕️ Techpresso!