today's ai news covers legal disputes, big funding rounds, and new tools. openai may sue apple over a chatgpt deal that fell short. recursive superintelligence raised $650 million for self-improving ai. cerebras had a strong ipo, and ibm released compact multilingual embedding models. other stories include benchmark audits, hardware gadgets, and job cuts at cisco.

  1. openai considers legal action against apple over chatgpt deal - this shows the high stakes and tensions in big tech ai partnerships when subscriber and revenue goals are missed.
  2. ai that builds itself gets $650m - recursive superintelligence's funding signals investor belief in ai that can find and fix its own flaws without human help.
  3. cerebras raises $5.5b in ipo - the chipmaker's ipo at a $56.4 billion valuation marks a strong start for tech listings in 2026.
  4. granite multilingual r2: compact 97m model tops sub-100m retrieval - ibm's open-source embedding models cover over 200 languages and code, making multilingual ai more accessible.
  5. benchjack audits ai agent benchmarks for reward hacking flaws - automated detection of reward hacking in benchmarks reveals widespread vulnerabilities that could mislead ai progress.
  6. cisco cuts 4,000 jobs to fund ai and security - even with record revenue, cisco is shifting resources toward ai and cybersecurity, reflecting industry-wide reprioritization.

other notable updates include a tiny desktop gadget for tracking claude token usage, a study on aligning ai confidence with human self-confidence to speed up learning, and a $23 million raise for wirestock's pivot to ai data supply. legal tech also saw movement with clio hitting $500 million in annual recurring revenue.