AI News for April 1, 2026 — 7min.ai

[NEWS] Leaked Claude Code exposes $2.5B ARR, KAIROS agent, and traps to poison AI copycats: Analysis of the leaked ~512,000 lines of TypeScript revealed concrete details not previously public: Claude Code generates $2.5B ARR; a codenamed KAIROS always-on background agent; anti-distillation defenses that inject fake tools to poison models attempting to copy Claude Code's behavior; and frustration-detection regexes.
[NEWS] OpenAI closes $122B round at $852B valuation, discloses $2B/month revenue and 900M+ weekly users: The round closed at $122B, co-led by SoftBank, Andreessen Horowitz, D.E.
[NEWS] Disney learned of Sora shutdown less than an hour before the public—and its OpenAI partnership is now 'effectively dormant': A WSJ investigation revealed Disney had an enterprise Sora pilot and learned of the shutdown less than an hour before OpenAI's public announcement.
[NEWS] Nvidia invests $2B in Marvell, deepens AI chip partnership via NVLink Fusion: Nvidia announced a $2B investment in Marvell Technology alongside a strategic partnership through NVLink Fusion, enabling Marvell to connect its silicon photonics products to Nvidia's AI infrastructure.
[TOOLS] Alibaba releases Qwen3.5-Omni, omnimodal AI beating Gemini 3.1 Pro on audio across 74 languages: Alibaba released Qwen3.5-Omni, an omnimodal AI model processing text, images, audio, and video natively, pretrained on 100M+ hours of audiovisual data.
[TOOLS] Slack adds 30+ AI features to Slackbot, gains MCP support and desktop agent capabilities: Salesforce unveiled 30+ new capabilities for Slackbot, transforming it into a full enterprise agent that takes meeting notes across video providers, operates on desktop outside Slack, and executes tasks via Model Context Protocol.
[NEWS] CoreWeave raises $8.5B in largest-ever chip-backed debt deal: CoreWeave secured $8.5B from banks and investors to finance cloud computing expansion, calling it the largest chip-backed debt deal of its kind.
[NEWS] Axios npm supply chain attack installs RAT on 135+ systems in 3-hour window: Attackers stole a long-lived npm token from axios's lead maintainer and published two poisoned versions that dropped a cross-platform remote access trojan.
[RESEARCH] Meta's semi-formal reasoning technique boosts LLM code review accuracy to 93%: Meta researchers introduced "semi-formal reasoning," a structured prompting method that requires AI agents to fill out a logical certificate with explicit premises, execution paths, and formal conclusions before answering.
[NEWS] OpenClaw's 500,000 instances lack enterprise kill switches, CEO's instance sold on dark web: A UK CEO's OpenClaw instance was listed on BreachForums for $25,000, offering root access, full AI conversation history, production databases, API keys, and personal family details.

About 7min.ai · Subscribe to newsletter