OpenAI shipped GPT-5.5 on April 23, just one week after Anthropic's Opus 4.7, positioning it as a step toward a unified AI "super app." GPT-5.5 matches Opus 4.6 per-token latency while delivering major gains in agentic coding, computer use, and long-context work — claiming leads on Terminal-Bench 2.0, BrowseComp, and OSWorld, though Opus 4.7 still leads on SWE-bench Pro and GPQA. Priced at $5/$30 per 1M input/output tokens with a 1M context window.
Google will invest $10B upfront in Anthropic and up to $30B more contingent on performance milestones, just days after Amazon pledged $25B. The deal includes 5 gigawatts of compute capacity over five years via Google Cloud TPUs. This is the largest single investment in an AI company to date and ends the narrative of a clean "Big Three" — OpenAI is now the odd one out without a hyperscaler parent.
Anthropic's implied valuation on Forge Global reached the $1 trillion mark this week, nearly tripling from its $380B Series G just three months ago. The surge is driven by annualized revenue leaping from $9B (end of 2025) to $30B+ (March 2026). Anthropic now trades above OpenAI ($880B) on private markets — though secondary prices carry significant liquidity discounts and caveats.
Anthropic shipped Claude Design alongside Opus 4.7, a tool that turns text prompts into interactive prototypes through conversational refinement, inline comments, and AI-generated adjustment sliders. CPO Mike Krieger resigned from Figma's board days before the launch. Figma shares fell ~7% on the news, signaling the market sees this as a credible threat to the $20B design tool category.
The Information reported that Google co-founder Sergey Brin has assembled a dedicated team within DeepMind, led by pre-training engineer Sebastian Borgeaud, to close the widening gap with Anthropic on agentic coding tasks. Internal data reportedly shows Google's AI code-writing rate at ~50% versus near-100% for Anthropic. Brin wrote in a staff memo: "We must urgently bridge the gap in agentic execution."
A White House OSTP memo accused China-based entities of running "deliberate, industrial-scale campaigns" to distill U.S. frontier AI systems, using tens of thousands of proxy accounts and jailbreaking techniques to extract proprietary capabilities from models built by OpenAI and Anthropic. Beijing denied the claims. The accusation lands three weeks before a Trump-Xi summit scheduled for May 14.
Meta announced it will lay off 10% of its workforce (roughly 8,000 employees) starting May 20, while canceling 6,000 open roles. The cuts fund AI capex projected at $115–135B in 2026, up from $72B in 2025. Zuckerberg framed it plainly: "Projects that used to require big teams are now accomplished by a single very talented person" with AI.
Over 103,000 AI agents were built by Defense Department employees in under five weeks using Google Gemini's no-code Agent Designer on GenAI.mil, with 1.1 million agent sessions logged. The tool is accessible to all 3 million DoD employees for unclassified work. The Pentagon plans to add OpenAI and xAI models to create a multi-model environment.
VP of Science Kevin Weil, Sora lead Bill Peebles, and B2B Applications head Srinivas Narayanan all departed OpenAI on April 17 as the company shuttered its money-losing Sora consumer app and consolidated around core products ahead of a potential IPO. This follows medical leaves by product chief Fidji Simo and marketing chief Kate Rouch, and Brad Lightcap's shift to "special projects."
Security researchers identified a convincing fake Claude download site that installs the real Claude app as a front while silently deploying PlugX malware via DLL sideloading. A separate campaign on GitHub used a fake Claude Code source leak to distribute the Vidar infostealer and GhostSocks proxy malware. With Claude approaching 290 million monthly web visits, it has become a high-value impersonation target.
