AI News for May 2, 2026 — 7min.ai

[NEWS] Pentagon signs classified AI deals with 8 tech giants, freezes out Anthropic: The Department of War announced agreements with SpaceX, OpenAI, Google, Nvidia, Reflection, Microsoft, AWS, and Oracle to deploy AI across classified networks at Impact Levels 6 and 7.
[NEWS] Meta acquires Assured Robot Intelligence to power humanoid push: Meta acquired San Diego startup Assured Robot Intelligence (ARI) for an undisclosed sum, folding the ~20-person team into Superintelligence Labs to build foundation models for humanoid robots.
[NEWS] Apple raises Mac Mini base price to $799 as AI demand drains supply: Apple pulled the $599 256GB Mac Mini and reset the starting price to $799 for the 512GB M4 model, just as Tim Cook warned on the Q2 call that the Mac Mini and Mac Studio 'may take several months to reach supply-demand balance.' Cook attributed the surge to buyers running agentic AI tools locally, with customer recognition arriving 'faster than what we had predicted.' Apple plans to bring Mac Mini production to the US later this year.
[NEWS] 200,000 MCP servers exposed by command-execution flaw Anthropic calls a feature: OX Security researchers found that MCP's STDIO transport, the default for connecting AI agents to local tools, executes any operating system command it receives without sanitization or boundary between configuration and command.
[NEWS] xAI ships Grok 4.3 with steep price cuts, launches Imagine agent mode: xAI released Grok 4.3 at $1.25 per million input tokens and $2.50 per million output tokens, undercutting OpenAI and Anthropic's flagships.
[RESEARCH] GPT-5.5 matches Claude Mythos on UK AI Security Institute cyber tests: The UK's AI Security Institute reported GPT-5.5 hit 71.4% on Expert-level capture-the-flag challenges, edging Mythos Preview's 68.6%.
[TOOLS] Anthropic launches Claude Security to give defenders an AI edge: Claude Security entered public beta for Claude Enterprise customers, scanning code for vulnerabilities and recommending patches using Claude Opus 4.7.
[NEWS] Linux 'Copy Fail' privilege escalation found via AI scan in one hour: Theori security researchers disclosed CVE-2026-31431, a Linux kernel flaw in the crypto subsystem that lets any user gain root via a single Python script across nearly every distribution released since 2017, with no per-distro offsets or recompilation.
[NEWS] Uber burns entire 2026 AI budget on Claude Code in four months: Uber's CTO disclosed the company spent its full 2026 AI budget by April after rolling out Claude Code in December 2025, with usage doubling by February.
[NEWS] Replit's Amjad Masad on Cursor deal, fighting Apple, billion-dollar run rate: Replit grew from $2.8M revenue in 2024 to a billion-dollar annual run rate, with net revenue retention 'as high as 300%,' CEO Amjad Masad said at TechCrunch StrictlyVC.

About 7min.ai · Subscribe to newsletter