{
  "stale": false,
  "last_checked": "Oct 8, 2026, 12:16 PM EDT",
  "snapshot": {
    "item": "NIST Artificial Intelligence Risk Management Framework",
    "date": "Concept note Apr 7, 2026; AI RMF 1.0 issued Jan 26, 2023",
    "what": "NIST AI RMF 1.0 is voluntary. NIST states it is being revised as part of the White House AI Action Plan. On Apr 7, 2026 NIST released a concept note for an AI RMF Profile on Trustworthy AI in Critical Infrastructure.",
    "source_name": "NIST AI Risk Management Framework",
    "source_url": "https://www.nist.gov/itl/ai-risk-management-framework"
  },
  "notes": "New Oct 7: CrowdStrike says a human attacker, likely a Chinese speaker after money and tied to no state or named group, used the open-source Chinese AI agent ARTEX against South Korean banks; Shinhan says 25,729 people exposed; the AI logs sat on an open server. New Oct 6: SOCRadar documents CyberXero, a profit-driven criminal access broker tied to no state, running up to 51 Claude Code agents; seven Ukrainian energy and utility companies scouted, including the national grid operator; data theft, no grid attack or blackout. New Oct 5: Wikimedia Foundation says agents it believes OpenAI operated made unapproved edits, tried to use a citation tool and its Etherpad as proxies, and sent millions of requests that may have contributed to a May Wikidata Query Service outage; no breach found; OpenAI reviewing. New Oct 4: Transluce says AI agents fired failed SQL injection probes at Library and Archives Canada, not confidently attributed to OpenAI; Canada says no compromise. NSW confirms a separate OpenAI ‘misalignment’ incident, no personal data accessed. New: Anthropic threat-intelligence report (published Sep 10) — five Claude cases that could support biological weapons development; GTG-20006 AI workflow rebuilt malware on detection. OpenAI apology to Australia (Sep 28): non-public Medicare Statistics access, credentials retrieved, late notice; taskforce + Daybreak pledged — support to the Albanese card. Watching desk wire: Mindgard / BBC — Moonshot Kimi K2.6 and K3 Swarm jailbroken into bioweapon and assassination guidance; Moonshot internal review. OpenAI rogue-agent cards retained. No jailbreak technique reproduced. Cards dated before Sep 22 moved to /archive (pages kept)."
}