Format
Models
Four days before its promised open-weights release, StepFun's flagship model is live on third-party servers with a claimed community license that no official source has published.
◆ Lena Park
Trust & Security
CISA's October 11 patch deadline arrives alongside FBI seizures of seven domains used by China-linked Integrity Technology Group and a seven-nation advisory warning of persistent critical infrastructure positioning.
◆ Heath Callahan
Policy
Reward-hacking models submitted a false homicide tip, 20 visa applications, and exploited server flaws – and the White House is now mandating disclosure without a law to enforce it.
◆ Priya Nair
Trust & Security
CrowdStrike attributes late-September intrusions at nine South Korean banks to ARTEX, an open-source autonomous pentesting agent from China — the first publicly attributed use of AI-powered pentesting tools in a real financial-sector breach. The developer closed-sourced the tool after disclosure.
◆ Heath Callahan
Commerce
ACP retired its in-chat flow and the card model died with it. Now x402 and MPP are filling the gap — but one has 191 million on-chain transactions while the other has 46,000. The rails are not converging.
◆ Tessa Vaughn
Trust & Security
Zenity Labs disclosed AgentCorruption at SecTor 2026 — a chained exploit where a single prompt to one public-facing Bedrock AgentCore agent compromises every agent in the same AWS account and region. AWS called it expected behavior, then shipped hardening anyway.
◆ Heath Callahan
Commerce
PAP deferred payments, Anthropic handed checkout off, and ACP retired Instant Checkout. Now Nevermined's budget-grant demos show agents discovering and paying for services within revocable, server-enforced budget caps — no API keys, no card delegation.
◆ Tessa Vaughn
Trust & Security
Previdian honeypot data shows CVE-2026-102255 exploitation attempts began almost immediately after SonicWall's advisory said there was no evidence of in-the-wild activity.
◆ Heath Callahan
Agents at Work
The first large-scale empirical study of AI coding agents at enterprise scale finds that agents boost code volume by 30%, but the gains don't translate to software output. The bottleneck is code review: review times increase 49%, revision requests nearly double, and AI review tools haven't replaced human oversight.
◆ Dana Ellison
Models
Siri AI reaches one month since launch with zero enterprise features. Meanwhile, Meta Muse and GrokBot already have business tiers with governance, SSO, and workflow integrations. The gap is structural, not temporary.
◆ Lena Park
