Daily AI RoundupsSunday, 20 September 2026

StepFun announces Step 5 Preview with 600B total parameters, 27B active per token and a one-million-token context window, with open weights planned for October 15. The Financial Times reports up to $300 billion in AI infrastructure guarantees over the past year, while Trump proposes an AI Force and a new AI czar without implementation details. Google releases AX v0.3.0 for declarative agent orchestration, and CXMT says its G5 DRAM process reaches mass production with at least 50% more dies per wafer. Nathan Lambert examines the limits of recursive self-improvement, Keyvan challenges AI hacking coverage, and Jonathan Kanter argues for liability and competition rather than an antitrust exemption.

• AI models
• AI infrastructure
• AI policy
• AI agents
• memory chips
• AI safety

Stories in this roundup

1. 01
   StepFun announces Step 5 Preview with 600B total parameters, 27B active and 1M context; open weights planned for October 15StepFun
2. 02
   Trump says he plans an AI Force and a new AI czar, without detailing how either would workReuters
3. 03
   FT reports Big Tech has offered up to $300B in AI infrastructure guarantees over the past year, with little exposure on balance sheetsFinancial Times
4. 04
   Nathan Lambert argues AI self-improvement is likelier to cut costs than trigger runaway intelligence, citing scaling costs and research bottlenecksNathan Lambert
5. 05
   Keyvan argues autonomous-AI hacking reports omit attack instructions and failures to isolate security tests from the internetKeyvan
6. 06
   Former DOJ antitrust chief Jonathan Kanter argues AI safety needs company liability and competition, not an antitrust exemptionThe Verge
7. 07
   Google releases AX v0.3.0, redesigning its agent orchestrator around declarative tasks, Redis and sandboxed workersGoogle
8. 08
   CXMT says its G5 DRAM process enters mass production, delivering at least 50% more dies per wafer and new 24Gb LPDDR5X chipsReuters