Tech policy, AI supply chains, and security: late-February/early-March 2026 snapshot

Late February reporting centered on a White House directive to halt federal use of Anthropic technology following Pentagon concerns, alongside renewed focus on AI compute supply as Meta and AMD were reported to have reached a very large chip agreement. In parallel, cybersecurity reporting highlighted how quickly state-aligned operators can weaponize patches, with Russian-state hackers exploiting a newly fixed Microsoft Office flaw within days. Product-level updates from Google and Samsung added consumer and platform context, including Google I/O timing and the shutdown of a consumer security feature.

Policy and procurement pressures around advanced AI systems sharpened in late February after multiple outlets reported the same core development: President Trump directed federal agencies to immediately stop using Anthropic’s technology. Reuters framed the decision as following a Pentagon determination that the startup represented a supply risk, and Defense One, The Hill, and CBS News each reported the order to “immediately cease” usage. Taken together, the coverage points to a rapid shift in federal posture toward at least one major AI provider, with the Pentagon’s supply-chain assessment presented as the key trigger.

At the same time, AI compute sourcing remained a central theme in corporate supply chains. The Financial Times reported that Meta agreed a multibillion-dollar chip deal with AMD. The Wall Street Journal went further on scale and structure, reporting that Meta and AMD agreed to an AI chips deal valued at more than $100 billion and that the arrangement included a warrant structure tied to delivery milestones. While the two accounts align on the existence of a significant AMD-Meta agreement, they differ on the magnitude and the detail available publicly, leaving some uncertainty about how the deal’s headline value is defined and how much of the structure is confirmed outside the Journal’s reporting.

Additional supply-chain claims circulated from a less-established source. Silicon Analysts claimed Intel secured $7 billion in strategic investments, including $5 billion from Nvidia and $2 billion from SoftBank, aimed at accelerating Intel’s IDM 2.0 and foundry roadmap. The same piece also claimed that demand for traditional server chips driven by AI data-center build-outs was pushing server CPU lead times to roughly 30 to 36 weeks. These points speak to the broader narrative of capacity constraints spilling beyond GPUs into general server components, but they are presented as claims from an analysis site rather than corroborated reporting in the provided source set.

Cybersecurity reporting underscored how quickly the threat environment can turn routine patching into an urgent operational race. Ars Technica reported that Russian-state hackers exploited a critical Microsoft Office vulnerability tracked as CVE-2026-21509 in less than 48 hours after Microsoft issued an urgent, unscheduled update. According to the report, researchers said the campaign compromised devices in diplomatic, maritime, and transport organizations across more than half a dozen countries. Ars Technica also reported that the attackers reverse-engineered the patch and developed an exploit that installed one of two previously unseen backdoor implants. The episode illustrates a continuing pattern in which high-value actors operationalize newly released fixes quickly, compressing the window defenders have to deploy patches before exploitation begins.

Consumer and platform updates added a separate layer of near-term context. TechCrunch reported that Google would discontinue its “dark web report” feature starting February 16, 2026, and said Google attributed the shutdown to feedback that the feature “didn’t provide helpful next steps.” The Verge reported Google announced dates for I/O 2026 and said the event would feature its latest AI breakthroughs and product updates, signaling that Google intends to keep AI positioning central to its flagship developer event.

On the hardware side, the Times of India reported Samsung’s Galaxy Book 6 series was set to launch in the United States on March 11 with a starting price of $1,049. The outlet reported Samsung said the lineup includes the Galaxy Book 6 Ultra, Galaxy Book 6 Pro, and Galaxy Book 6, and that the devices are powered by Intel Core Ultra Series 3 processors built on Intel’s 18A process. Together, these product notes sit adjacent to the supply-chain storylines: while governments and hyperscalers contend with sourcing, consumer-facing vendors continue to market new systems and timelines that implicitly depend on stable semiconductor availability.