Daily Tech Briefing: Meta's Wearable AI Bet, Muse Agent & Google's Orbital Data Center

Daily Tech Briefing | September 25, 2026

AI Agents Go Mainstream — and So Do Their Security Risks

Good morning, folks! Erwin here, back with today’s biggest technology stories. Meta used its annual Connect conference to go all-in on consumer AI agents and wearable devices, unveiling the Muse agent as the "centrepiece" of its AI strategy while CEO Mark Zuckerberg explicitly rejected calls for a coordinated industry slowdown. 

Google announced plans to launch its first experimental satellite next week to test whether AI chips can operate reliably in space — a step toward solar-powered orbital data centers. AI infrastructure startups Verda and Nava are in advanced funding talks at valuations of $1 billion and $750 million respectively. And two separate security disclosures this week confirmed that autonomous AI agents are now actively breaching enterprise systems: Spain's data protection agency reported the country's first AI-agent-driven data breach, while Cloud Security Alliance research documented a campaign in which a single human operator used three open-source LLM tools to breach more than 100 retailers and steal 600,000 payment cards. The through-line is unmistakable: AI agents have moved from demo to deployment — and the security, infrastructure, and governance frameworks around them are racing to catch up.

1. Meta Bets Big on Wearable AI and the Muse Agent

Meta is betting that consumer adoption of agents and wearables will drive AI's next phase, not enterprise deployment alone.

What happened: At Meta's Connect conference on Thursday, Zuckerberg opened his keynote by acknowledging that "it's no secret that people in the industry have a wide range of views about how the future is going to go," but positioned Meta as an optimist. "We believe that the best way to create a positive future for everyone is to put power in people's hands," he said. "The path to a safe future is to distribute this technology widely and create the right checks and balances, rather than concentrating it in the hands of just a few."

The centerpiece of the announcement was Muse, Meta's AI agent, demonstrated as a consumer-friendly assistant capable of real-time voice and video conversations, travel planning, restaurant booking, grocery ordering, and other tasks. Meta has partnered with OpenTable, Spotify, Ticketmaster, and others to enable these integrations. Meta plans to bring Muse to its AI glasses lineup for hands-free access, and will release a keychain-like device called Muse Charm in December that packs the "whole Muse experience" into a portable form factor.

Key numbers/companies: Muse agent · Muse Charm (December) · Ray-Ban Meta Audio at $349 (October) · Partners: OpenTable, Spotify, Ticketmaster · Meta.

Meta also launched Ray-Ban Meta Audio, an audio-only version of its smart glasses with no screen and no video recording capability — a direct response to privacy concerns about users recording others without permission.

Why it matters: Zuckerberg's framing — "empowering people is the source of prosperity" and "the highest purpose of superintelligence is creation and invention, not automation" — is a direct counter to the slowdown arguments advanced by Anthropic CEO Dario Amodei earlier this month. The Muse Charm and audio glasses signal that Meta sees form-factor diversity, not a single device, as the path to mass adoption. The question is whether privacy safeguards and agent reliability can keep pace with the company's distribution ambitions.

2. Google Plans Orbital Data Center Experiment

The AI infrastructure race is no longer confined to Earth — but the engineering challenges are formidable.

What happened: Google announced it will launch its first experimental satellite next week to test whether AI semiconductors can operate reliably in space, where they face intense vibration during launch, strong radiation, and the challenge of continuous solar power management. The experiment is a step toward building solar-powered AI data centers in orbit — a concept gaining attention as terrestrial data centers face mounting power consumption constraints. SpaceX is pursuing a similar initiative.

Key numbers/companies: First experimental satellite launch next week · AI semiconductor radiation and thermal testing · Google, SpaceX.

Why it matters: Orbital data centers would offer near-continuous solar power, eliminate land and water constraints, and potentially reduce latency for certain applications. But the engineering challenges remain: radiation hardening, thermal management in vacuum, and the cost of launching and maintaining hardware in orbit. Google's satellite is an early experiment, not a deployment — but it signals that the industry is beginning to think beyond terrestrial constraints for the next generation of AI compute.

3. AI Infrastructure Startups Attract Major Funding

Investors see infrastructure — not just models — as the durable value layer, across geographies, asset types, and capital structures.

Three significant funding developments this week underscored investor appetite for AI infrastructure at every layer of the stack.

Verda completed a $189 million financing round led by Emergence Capital, with participation from MUFG Innovation Partners, Super Micro Computer, Varma Mutual Pension Insurance, and Lifeline Ventures. The company's post-financing valuation reached at least $1 billion.

Nava, an AI data center startup founded by former OYO COO Abhinav Sinha, is in talks to raise approximately $200 million from Greenoaks at a valuation of around $750 million. Nava builds private cloud infrastructure for enterprises, offering a secure and scalable stack designed for the AI era. If completed, it would rank among the largest funding rounds in India's AI and data center ecosystem, behind only Sarvam AI's $300 million raise.

USD.AI, an on-chain lending protocol, secured a $128.9 million GPU financing facility — the largest single financing in the protocol's history — to support an undisclosed borrower deploying 32 Nvidia GB200 NVL72 systems in British Columbia, Canada.

Additionally, Island, an AI startup, raised $400 million in Series F funding at a $6.4 billion valuation, led by Evolution Equity Partners with participation from Prysm Capital, Sequoia Capital, Coatue Management, Cyberstarts, Insight Partners, and JP Morgan Growth Equity Partners.

Why it matters: These deals illustrate that AI infrastructure financing is diversifying across geographies, asset types, and capital structures. The GPU financing facility in particular shows that asset-backed debt markets are maturing to underwrite specific hardware deployments.

AI Infrastructure Funding Deals

Company	Raise / Valuation	Strategic Significance
Verda	$189M round; ~$1B valuation	Emergence Capital lead; Super Micro participation signals hardware-financing convergence
Nava	~$200M (in talks); ~$750M valuation	India private cloud for AI; one of largest rounds in India's AI ecosystem
USD.AI	$128.9M GPU financing facility	32 Nvidia GB200 NVL72 systems; asset-backed GPU debt maturing
Island	$400M Series F; $6.4B valuation	Enterprise AI platform; Sequoia, Coatue, Insight participation

4. AI Agent Security: From Theoretical to Operational

The write path — not read access — is where the risk concentrates. Anything that changes state should remain a human-approved step.

What happened — Spain's first AI-agent breach: Spain's data protection agency (AEPD) reported that it received its first notification of a personal data breach allegedly carried out using an AI agent. According to the notification, the agent searched for vulnerabilities in generic files, successfully logged in, and autonomously examined the application until it located a flaw that allowed it to modify personal data and access invoices. The affected organization has not been disclosed, nor has the number of people affected. AEPD is continuing to analyze the incident.

What happened — 100+ retailers breached: Cloud Security Alliance research documented a campaign in which a financially motivated operator used three open-source, LLM-driven tools to autonomously breach more than 100 e-commerce sites and exfiltrate over 600,000 payment card records between July and September 2026, at an average cost of roughly $25 per target. The operator issued fewer than 2,000 short prompts across 260 sessions to sustain 105 distinct attack waves.

The operator relied on three open-source tools: Strix (penetration-testing framework), Cairn (autonomous exploitation engine), and Hermes (orchestration layer). For decision-making, the operator used Anthropic's Claude Opus 4.6 through OpenRouter, reportedly after finding that newer models declined to carry out the requested actions.

Why it matters: These two incidents — one a regulatory disclosure from Spain, the other a detailed reconstruction of an attacker's own staging server — provide the most concrete evidence yet that agentic AI is being weaponized at scale. The economics are striking: $12,000–$18,000 in total campaign cost for over 600,000 stolen card records. Enterprises must now treat AI agents as first-class security principals, subject to the same identity, permission, and audit controls as human users.

AI Agent Security Incidents & Responses

Incident	Scale / Tooling	Enterprise Response
Spain AI-agent data breach	Autonomous login, data modification, invoice access	AEPD policy update: add agents to risk analysis; tighten credentials
100+ retailer breach	600K cards; ~$25/target; Strix, Cairn, Hermes; Claude Opus 4.6	Treat agents as machine identities; state-change approval gates

5. Cloud Providers Diverge as AI Demand Accelerates

Demand is not the constraint — supply is. Multi-cloud strategies may be necessary to access the compute enterprises need.

What happened: Bernstein's latest report on hyperscale cloud providers shows significant divergence in Q2 2026 growth rates: Google Cloud grew 82%, Oracle OCI surged 121%, Alibaba Cloud grew 44.9%, Microsoft Azure grew 43%, and Amazon AWS grew 37%.

Key numbers/companies: Google Cloud +82% · Oracle OCI +121% · Alibaba Cloud +44.9% · Azure +43% · AWS +37% · Azure revenue crossed $100B · Microsoft order backlog $678B · AWS 2026 capex $200B.

Why it matters: The growth divergence reflects different strategies. Google Cloud is benefiting from its TPU ecosystem and AI-native positioning. Oracle's OCI is growing from a smaller base but gaining traction with AI workloads. Azure's backlog of $678 billion signals that demand is not the constraint — supply is. AWS's $200 billion capex commitment shows that the company is willing to invest aggressively despite slower growth. For enterprises, the implication is that cloud AI capacity remains constrained.

6. AI Governance: Self-Regulation and the "Middle Path"

The industry is moving toward self-regulation while regulators push for international cooperation.

What happened: Google, OpenAI, and Anthropic are reportedly in discussions to establish a self-regulatory body for frontier AI safety standards, tentatively named the "Standard for Advanced Frontier AI" (SAFA). The organization would operate independently without government regulation, with a target launch by the end of 2026 or early 2027.

Separately, OpenAI CEO Sam Altman called for a "middle path" on AI regulation, arguing that the industry should neither be blindly optimistic nor fall into doom scenarios. He emphasized the need to ensure AI remains under human control. President Trump rejected a "globalist framework" for AI governance on Tuesday.

Why it matters: The SAFA initiative is a direct response to the slowdown debate and to growing regulatory pressure in the EU and elsewhere. By creating a self-regulatory body, the leading AI labs aim to preempt government mandates while establishing shared safety standards. But the effectiveness of self-regulation depends on enforcement — and on whether the standards are rigorous enough to satisfy regulators. Altman's "middle path" framing attempts to bridge the gap between accelerationists like Zuckerberg and cautionists like Amodei. The Trump administration's rejection of a globalist framework adds another layer of complexity: U.S. policy is likely to favor industry-led approaches over international agreements.

What to Watch Next

Company / Topic	Upcoming Catalyst
Meta	Muse agent rollout; Muse Charm launch in December; Ray-Ban Meta Audio availability in October
Google	Experimental satellite launch next week; orbital data center feasibility data
Verda / Nava	Finalization of funding rounds; enterprise customer announcements
AI agent security	Enterprise adoption of agent identity and permission controls; AEPD investigation findings
Cloud providers	Q3 earnings commentary on AI capacity constraints and backlog trends
SAFA	Formal launch timeline; participation scope; standards framework details
Nvidia	H200 China delivery progress; competitive response to Google TPU momentum

Related CODEW Coverage

• AI Infrastructure Watch — Data-center economics, GPU financing, and cloud capacity
• Cybersecurity Watch — AI agent security, machine identity, and data-layer controls
• Startup Funding Watch — Deal structures and capital flows in AI infrastructure
• Cloud Computing Watch — Sovereign AI, hyperscaler competition, and regional data residency
• AI Watch — Agent platforms, model releases, and consumer AI adoption

SOURCES & REFERENCES

Meta Connect: Waikato Times / AP — "Meta bets big on wearable AI, new AI agent as Zuckerberg pushes back against AI doom" (Sept. 25, 2026).

Google Orbital Data Center: Tokyo News / Kyodo — "Google to launch first experimental satellite for space data center" (Sept. 25, 2026).

AI Infrastructure Funding: ChainCatcher — "Verda completed a $189 million financing round, led by Emergence Capital" (Sept. 24, 2026) · Moneycontrol — "AI data centre startup Nava, founded by ex-OYO executive, in talks to raise $200 million from Greenoaks" (Sept. 24, 2026) · BlockBeats — "USD.AI Secures $128.9 Million GPU Financing Facility" (Sept. 24, 2026) · ChainCatcher — "Island completes $400M Series F at $6.4B valuation" (Sept. 24, 2026).

AI Agent Security: INCIBE — "Personal Data Breach Reported in Spain Linked to an Attack Carried Out by an Artificial Intelligence Agent" (Sept. 24, 2026) · Cloud Security Alliance — "Autonomous AI Agents Breach 100+ Retailers: Security Implications" (Sept. 24, 2026).

Cloud Providers: KuCoin / Bernstein Report — "Hyperscale cloud providers diverge in Q2; Google leads with 82% growth" (Sept. 24, 2026) · CINN — "AI compute race escalates: global cloud vendors accelerate data center and full-stack capabilities" (Sept. 24, 2026).

AI Governance: Ming Pao — "OpenAI CEO calls for 'middle path' to ensure AI is controlled" (Sept. 25, 2026) · Tokyo News / Kyodo — "Google, OpenAI, Anthropic considering new self-regulatory AI safety body" (Sept. 25, 2026) · Cailian Press — "Google, OpenAI, Anthropic reportedly plan to jointly establish 'Standard for Advanced Frontier AI' (SAFA)" (Sept. 25, 2026).

THE CODEW · DAILY TECH BRIEFING

Editorial Note

The CODEW Daily Tech Briefing is a fast morning read on the day's most important technology signal, plus the handful of other stories worth knowing — built to be read in minutes, with the deeper analytical work reserved for The CODEW's Watch series and Weekly Tech Roundup.

Coverage is based on company announcements, public disclosures, industry reporting, and other publicly available information. Reported figures and sourced-but-unconfirmed details are noted as such. Analysis reflects the reporting period and should be considered in the context of the sources and developments cited.