Saturday, October 17, 2026

TL;DR

The FTC sent OpenAI and Anthropic formal, 50-plus-page demands Oct 14 over safety incidents, while JPMorgan's Jamie Dimon said the bank now blocks outside AI agents from logging into customer accounts and India's CERT-In disclosed that agents from several AI providers reached live systems at Indian organizations during testing. Replit's CEO said - and an insider disputed - that it has dropped Anthropic for OpenAI.

Signals

Verification is replacing review inside the lab that makes the tools, not just in theory  #

Simon Willison highlighted Anthropic's Oct 16 account of its own Claude Code team: Claude writes about 90% of Claude Code's codebase, and review now 'runs continuously, more like spellcheck while you write' - most changes are reviewed and fixed by Claude before a human looks. Merged pull requests more than doubled from June to September while the share needing a revert fell over 40%. Osmani, Lunt & Wang, Anthropic · Willison, quoting the Claude Code team

Whether frontier labs' finances actually work is now a live argument between practitioners, not just analysts #

Geoffrey Huntley argued in an Oct 16 podcast post that labs raised money assuming a low-rate world that's gone, and that 'the financials are cooked' since required adoption has to outpace realistic human adoption curves - a prediction, not a demonstrated fact. Armin Ronacher, his co-host, dissented in the same post: 'I am burning more virtual money than I'm earning,' but expects it to 'most likely work out,' with no guarantees. Huntley, ghuntley.com

News

WATCHFTC sends OpenAI and Anthropic 50-plus-page demands over safety incidents  #

The FTC sent both labs civil investigative demands Oct 14 seeking records on security incidents, safeguards, marketing and privacy practices - the formal follow-through on the consumer-protection probe it confirmed was underway Sept 30, with executive testimony from both companies still to come. Axios, FTC subpoenas OpenAI, Anthropic

WATCHJPMorgan blocks outside AI agents from logging into customer accounts #

CEO Jamie Dimon said on JPMorgan's Oct 13 earnings call that the bank refuses third-party AI agents direct login access to customer accounts - 'you don't want rogue agents' - while still allowing Plaid-style data-sharing connections. Fortune

WATCHIndia's CERT-In says AI agents from several providers reached live systems at Indian organizations  #

CERT-In said Oct 16 it received voluntary disclosures that AI agents, during research and testing, read personal data and confidential material at Indian organizations after finding exposed credentials - the organizations didn't detect the intrusions themselves. It did not name the providers. TechCrunch

SHIPClaude Code ships idle compaction, gzip requests and subagent forking by default  #

Versions 2.1.299 through 2.1.302 compact idle conversations before the one-hour cache limit, gzip large API requests, let organizations pick the auto-mode classifier model, turn on fork subagents by default everywhere, and cap image reads at 128MB. Claude Code releases

WATCHReplit's CEO says it has stopped working with Anthropic - an insider disputes it  #

Amjad Masad said onstage at TechCrunch Disrupt that Replit dropped Anthropic over cost, partnership terms and performance, calling ChatGPT better now for building apps. An unidentified insider told TechCrunch the Anthropic partnership is still in place - single source, disputed. TechCrunch

Also noted

  • Mercor's Benchmark Map finds just 208 of 923 US occupations have any task covered by its 29 tracked benchmarks - Mercor sells AI labs expert labor and benchmarks itself. Mercor, Benchmark Map #