On Monday afternoon, OpenAI and Anthropic announced rival joint ventures with private equity firms aimed at embedding AI engineers inside Fortune 500 companies. OpenAI's Deployment Company raised over $4 billion from TPG, Brookfield, Bain Capital and others, with a 17.5% guaranteed five-year return. Anthropic raised $1.5 billion alongside Blackstone, Hellman & Friedman, and Goldman Sachs. The deals are aimed at McKinsey, BCG, and Bain & Company.
On April 30, 2026, the Wall Street Journal reported that the White House blocked Anthropic's plan to expand Project Glasswing access from approximately 50 organizations to 120. The administration cited security concerns and compute capacity. The same administration continues to label Anthropic itself as a supply chain risk to the Department of Defense while drafting an executive action to allow federal agencies to onboard Mythos around the Pentagon's blacklist.
On April 30, 2026, attackers pushed two malicious versions of PyTorch Lightning to PyPI containing an 11MB obfuscated payload that stole GitHub tokens, npm tokens, SSH keys, AWS credentials, and Vault secrets on import. The packages were live for 42 minutes. The same TeamPCP threat group simultaneously compromised intercom-client and intercom-php, riding the dependency chain through pyannote-audio. ML teams need to rotate credentials and audit GitHub commits authored under impersonation identities.
Mistral AI released Mistral Medium 3.5 on April 30, 2026, a 128B dense model with a 256K context window that retires Magistral from Le Chat and replaces Devstral 2 in the Vibe CLI. The model is open-weights on Hugging Face under a modified MIT license, scores 77.6% on SWE-Bench Verified and 91.4 on the τ³-Telecom agentic benchmark, and self-hosts on four GPUs. API pricing is $1.50 per million input tokens and $7.50 per million output tokens. The release also introduced cloud-side Vibe Remote Agents that can run async coding sessions in parallel and a new Work mode in Le Chat for multi-step productivity tasks. The launch comes one month after Mistral borrowed 830 million dollars to build a 13,800-GPU data center outside Paris.
The U.S. Department of War announced on May 1, 2026 that eight AI companies (SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, Amazon Web Services, and Oracle) signed agreements to deploy advanced AI on the Pentagon's classified Impact Level 6 and 7 networks for lawful operational use. The list excluded Anthropic, which the Pentagon designated a supply chain risk in March after the company refused to remove safety guardrails preventing autonomous weapons targeting and domestic mass surveillance. A federal court in California granted Anthropic a preliminary injunction on March 26, but the D.C. Circuit Court of Appeals denied a stay on April 8, allowing procurement to proceed pending oral argument scheduled for May 19. The Pentagon's GenAI.mil platform now serves 1.3 million Department personnel.
Bloomberg reported on April 29, 2026 that Anthropic is reviewing preemptive offers to raise around $50 billion at a valuation of $850 to $900 billion, just 76 days after closing its Series G at $380 billion. A board decision is expected in May. An October IPO targeting $60 billion or more is still on the table.
Meta beat every Q1 2026 estimate with $56.31 billion in revenue and 33% growth, the fastest since 2021. The stock fell 10% anyway after CFO Susan Li raised the 2026 capex ceiling to $145 billion. On April 30, JPMorgan downgraded Meta from Overweight to Neutral, projecting capex would hit $202 billion by 2027 and push free cash flow into negative territory.
On April 27, 2026, China's National Development and Reform Commission issued a 54-character decree ordering Meta to unwind its over-two-billion-dollar acquisition of Manus, a Singaporean AI agent startup with Chinese roots. The deal closed in December 2025. Manus engineers and executives have already joined Meta. Founders Xiao Hong and Ji Yichao were summoned to Beijing in March and placed under exit bans. Bloomberg has called the corporate-relocation playbook used by Manus officially dead, signaling a new doctrine that exposes any cross-border AI acquisition to retroactive Chinese state intervention.
On April 28, 2026, Pentagon AI chief Cameron Stanley confirmed Google Gemini is now available on classified Defense Department networks for any lawful government purpose. The deal landed one day after more than 600 Google employees signed an open letter asking CEO Sundar Pichai to refuse. The same day, Google quietly exited a separate $100 million Pentagon drone swarm contest. The contract fills the void left when Anthropic was designated a supply chain risk in March after refusing classified work on principle.
Sam Altman apologized to the community of Tumbler Ridge, B.C., for OpenAI's failure to alert police after employees flagged the eventual shooter's ChatGPT account in June 2025, eight months before the February 10, 2026 mass shooting that killed eight people.
Microsoft and OpenAI announced on April 27, 2026 a sweeping restructuring of their partnership: Microsoft's license becomes non-exclusive through 2032, OpenAI is free to sell on AWS and Google Cloud, Microsoft stops paying revenue share, and the AGI trigger is replaced with a fixed calendar date.
DeepSeek V4-Pro hit 93.5 on LiveCodeBench and 80.6% on SWE-bench Verified Friday, ahead of Gemini 3.1 and within two-tenths of a point of Claude Opus 4.6. Open weights shipped under the MIT License on Hugging Face. Output tokens run $3.48 per million versus $25 for Opus 4.7.
Alphabet wired ten billion dollars at a $350 billion valuation Friday and pledged up to thirty billion more if Anthropic hits performance targets, locking in 5 gigawatts of TPU capacity over five years and up to one million Ironwood chips. Three days earlier, Amazon committed up to twenty-five billion of its own.
Microsoft announced its first-ever voluntary retirement program in 51 years on April 23, 2026. Roughly 8,750 US employees at senior director level or below are eligible under a formula that sums age plus years of service to at least 70. Formal notifications go out May 7, with 30 days to decide. The move comes as CEO Satya Nadella and CFO Amy Hood balance $80B+ AI infrastructure spending against shareholder pressure.
OpenAI released GPT-5.5 on April 23, 2026, six weeks after GPT-5.4. The new model posts state-of-the-art Terminal-Bench 2.0 (82.7%) and nearly doubles Claude Opus 4.7 on FrontierMath Tier 4, but API pricing doubled to $5 per million input tokens and $30 per million output tokens. A new GPT-5.5 Pro tier runs six times the standard rate.
Vercel disclosed on April 20, 2026 that a Context.ai OAuth token compromise gave attackers access to non-sensitive environment variables. The Context.ai breach started in February when an employee downloaded Roblox cheat scripts containing Lumma Stealer.
Anthropic's implied valuation on Forge Global hit $1 trillion, overtaking OpenAI's $880 billion on the same platform. The jump marks a 2.6x premium over Anthropic's $380 billion Series G from February 2026.
Bloomberg reported that members of a private Discord group gained access to Mythos Preview through a third-party vendor environment after identifying the model's online location. Anthropic said it was investigating the report and had found no evidence that its own systems were affected.
