Meta (META) on Solana

Amazon has blocked Meta's Muse AI agent from shopping on its site. On Sunday night, users sending Muse to buy on Amazon started getting a message that access by an unauthorized AI agent violated Amazon's Conditions of Use. GeekWire reports that Amazon's objection is that agents have to identify themselves in every request, and Muse did not. Amazon said third-party apps that buy on a customer's behalf "should operate openly and respect service provider decisions about whether or not to participate." Amazon has blocked agents from OpenAI and Google before. It sued Perplexity over its Comet browser agent last year, but the Ninth Circuit reversed the preliminary injunction in that case in August. Meta has not commented publicly on the block.

The dispute matters for Meta because Muse's business model depends on completing purchases across retailers. Mark Zuckerberg has said the agent would take "a very small cut" of transactions, and that fee could be paid by businesses rather than users. Muse's climb to the top of Apple's App Store built momentum for Meta ahead of its annual Meta Connect conference. According to GeekWire, Meta shares rose 11% on Monday to their highest close in a year. Being locked out of the largest U.S. online retailer limits what Muse can do for now. Analysts quoted by CNBC and GeekWire expect more fights like this as Apple, Google and possibly TikTok build their own personal shopping agents.

Meta Platforms stock completed a bullish triangle breakout on September 22, jumping 11% to close at $741.25 — its largest single-day gain since April 2025 — on volume several times the daily average. The move follows a 20%-plus rally since Muse, Meta's personal AI agent, launched on September 8 and quickly reached the top of Apple's U.S. App Store. Technically, META broke above resistance near $665 and cleared its 200-period moving average, with the triangle's measured objective pointing toward the $950 area. Wells Fargo raised its price target to $796 from $640, while the highest Wall Street forecast now sits at $1,000.

Chart technicians note the Stochastic oscillator has entered overbought territory, suggesting a short-term consolidation before any continuation. Key support on a pullback sits at the broken triangle boundary around $690, with the 200-period moving average at roughly $620 as a deeper floor. The analyst consensus average stands near $761, and active traders are watching the $700 level as a potential re-entry zone should the stock retrace before attempting a push toward the $900–$950 target range implied by the pattern.

Meta's Muse personal AI agent reached 2.5 million downloads in the 13 days following its September 8 launch, topping the U.S. iOS App Store free chart and briefly surpassing ChatGPT. In a 13-day post-launch comparison, Muse's cumulative downloads exceeded Anthropic's Claude (400,000) and xAI's Grok (200,000) by a wide margin; ChatGPT drew 3.1 million downloads over the same window after its own debut, still leading on an equivalent-period basis. Muse recorded approximately 730,000 downloads in its first five days alone.

The strong launch metrics reinforce the consumer reach of Meta's AI strategy, which differentiates Muse through deep integration across Facebook, Instagram, and WhatsApp rather than a standalone chat interface. That distribution advantage — reaching billions of existing Meta platform users without requiring a separate app switch — is seen as a key factor behind the rapid adoption and is directly material to Meta's monetization outlook for its AI investments.

Policymakers and technology leaders are debating the feasibility of AI kill switches — emergency mechanisms to shut down AI systems — and Meta Platforms is directly in the frame. CNBC reports that hyperscalers including Meta, Alphabet, and Amazon have poured billions into globally distributed data centers, making a centralized shutdown technically implausible. Tim Brown of cybersecurity firm Team8 captures the structural problem plainly: "There's not one entity to kill." Nick Warner, CEO of cyber startup Neo, adds that while a kill switch is "not too little," it is "probably too late" to serve as a panacea for AI's widening risks.

The debate carries specific regulatory weight for Meta given its dual role as both an AI infrastructure hyperscaler and the distributor of LLaMA, its open-source model family. Because LLaMA weights are publicly released and run on third-party hardware worldwide, any government-mandated shutdown mechanism would have no practical reach over deployed copies — a dynamic that regulators are increasingly noting. Experts cite AI unpredictability and agents' ability to circumvent controls as further arguments for graduated oversight frameworks rather than binary shutoff mechanisms, a policy direction that could impose new compliance requirements on Meta's AI roadmap.

Meta launched Muse on September 9, 2026 — an autonomous AI agent that can send emails, sell items, book travel, manage payments, and control smart-home devices on a user's behalf. Available through a dedicated app or WhatsApp, the agent is offered in three tiers (free, $20/month, and $100/month) and is currently limited to U.S. users. Built on the open-source OpenClaw system, Muse was delayed from April to September while Meta bolstered its security posture, including adding an "autonomous safety agent" to monitor Muse's actions in real time.

Internal testing nonetheless surfaced concerns — including an incident involving exposure of private iCloud photos — alongside reliability issues such as repeated logouts and inconsistent monitoring. Muse is central to CEO Mark Zuckerberg's "personal superintelligence" strategy and represents a push to reduce Meta's dependence on advertising revenue. The company has guided for AI infrastructure spending exceeding $130 billion in 2026.

Meta's in-house AI inference chip program now spans three shipped generations — MTIA v1 (Freya), v2 (Artemis), and v3 (Iris) — each paired with a dedicated server platform: Kings Canyon Gen2, Grand Teton-Artemis, and Santa Barbara Iris 1U respectively. New analysis of Meta's global MTIA rollout maps country-level deployments across nine countries in North America, Europe, and Asia Pacific, along with U.S. state and city-level datacenter footprint data, showing the scale at which Meta has replaced third-party accelerators for its AI inference workloads.

The report notes that Meta's evolving ASIC and server stack is generating supply-chain opportunities beyond the chips themselves, and flags MTIA v4, v5, and v6 as forthcoming in 2026. The multi-generational cadence mirrors the pace at which hyperscalers are iterating on custom silicon to reduce dependence on merchant GPU vendors for recommendation, ranking, and generative AI inference at scale.

Meta has acquired Stilla.ai, a Stockholm-based startup founded in 2024 that built a coordination layer connecting workplace tools such as Slack, Linear, GitHub, and Notion. The deal, whose terms were not disclosed, brings in co-founders Siavash Ghorbani and Kaj Drobin — former Shopify executives who built Shop and Shop Pay — along with the company's $5 million in pre-seed-backed technology. Rather than bolstering Meta's general-purpose AI assistants, the acquisition is aimed squarely at Meta Business Agent, an AI tool embedded across WhatsApp, Messenger, and Instagram that handles customer inquiries, product recommendations, appointment booking, and lead qualification for over one million businesses.

The move underscores Meta's growing emphasis on monetizing commerce through messaging rather than competing in the chatbot race. Meta's non-advertising revenue reached $1 billion in Q2 2026, up 73% year-over-year — outpacing the 27% growth in its core ad business. WhatsApp alone counts more than 200 million small business users globally, with paid messaging already generating an annualized $2 billion in revenue through WhatsApp Business Premium subscriptions and token-based pricing at $2.00 per million tokens. The Stilla.ai acquisition signals Meta's intent to make Business Agent a stickier, more capable commerce infrastructure layer across its family of apps.

A shareholder governance analysis argues that Meta's $17.1 billion child-safety settlement — covering claims by 47 states related to addiction harms, sexual abuse material, and trafficking exploitation on its platforms — was a foreseeable consequence of a structural accountability gap. The root cause cited is Meta's dual-class share arrangement: Mark Zuckerberg holds roughly 61% of voting power despite owning only about 13% of the company, because his shares carry ten votes each. When a 2021 shareholder resolution on content governance won 63.1% of the independent shareholder vote, Zuckerberg's bloc reduced the reported result to 19%, effectively vetoing a reform that two-thirds of financially exposed shareholders had backed.

The piece frames the settlement as a cost shareholders bore because one person could disregard repeated demands — from 2019 through 2021 — for platform safety improvements. Analysts note the payout may understate total liability, with plaintiffs' models suggesting potential damages in the trillions; Meta admitted no wrongdoing and the agreement includes only "best-effort" age verification commitments. Thousands of additional lawsuits remain pending, with trials set to resume in October, leaving ongoing exposure that independent shareholders have limited structural power to address.

Meta has finalized an $18 billion settlement with 29 state attorneys general over youth safety concerns on Instagram and Facebook, following a trial held in Oakland federal court in August. The agreement breaks into roughly $12.7 billion in unconditional payments over ten years, with a further $5.3 billion contingent on YouTube and TikTok adopting equivalent platform restrictions — including a two-hour daily use cap for minors, nighttime app blocking, and tighter age verification. The company expects to record approximately $10 billion in legal expenses in Q3 2026 as a result, a charge excluded from prior guidance. The settlement is substantially lower than the trillion-dollar liability that court filings had suggested was possible.

Morgan Stanley analysts framed the resolution as analogous to Google's antitrust settlement last year, which preceded a wave of successful AI launches, arguing it removes a significant legal overhang. Meta has not signaled any reduction in its $130–145 billion capital expenditure plan for 2026, the bulk of which is directed at AI infrastructure. The company is also preparing to launch Hatch, a consumer AI agent embedded in WhatsApp and Instagram capable of autonomous tasks such as purchases and restaurant bookings, with a rollout targeted for early September.