AI Governance & Risk

Under the Commission's GPAI guidelines, a fine-tune makes you a model provider only past one-third of the original training compute. LoRA never gets close; continued pretraining on a base with undisclosed compute can.

Anthropic's Claude Duet checks Gmail, Slack, HubSpot and QuickBooks every few hours from its own VM, but it launched only on personal Pro and Max plans with no admin controls. Your Workspace and Slack app policies are the only console you have.

Collibra bought trail ML on October 5 and has said nothing about standalone availability or trail's OneTrust and ServiceNow connectors. Export your evidence and put the integrations in the contract.

Zoomtopia 2026 made ZoomMate Agent Teams, MCP servers and the Search & Ask APIs available now, while Centralized AI Management, Compliance Manager for AI notes and AI Guardrails are dated December 2026 through Q1 2027. Here is the rollout order that follows.

ChatGPT's Chrome extension can now send messages and fill forms in any tab a user is signed into, on every plan. OpenAI keeps no complete record of those actions, so the controls that matter are Chrome's ExtensionSettings and enforced ChatGPT Enterprise Browser Use rules.

For published campaigns, buy Adobe Firefly on an enterprise contract and stay on Adobe's own models. Partner models, modified outputs and trademark claims fall outside most image generator indemnities, and Shutterstock does not publish its terms.

After the Otter wiretap ruling and the Granola class action, the safest AI note-taker for a large organisation is the one built into Teams, Meet or Zoom. Fireflies Enterprise is the standalone pick; Otter is the one to drop.

Meta's 816B Muse Spark Open is free for internal use, but its license bans operating critical infrastructure and any nuclear or military use, the same clauses Llama 3 carried.

Freshworks is buying Torii, a Gartner Leader in SaaS management, to put AI spend chargeback inside Freshservice. Torii customers on ServiceNow or Jira have no written continuity promise yet and should get one before the deal closes.

Workday is buying C1.ai (formerly ConductorOne) so HR events grant and revoke access for people and AI agents. The announcement says nothing about C1 customers whose HR system isn't Workday.

Colorado's ECDIS rules make your chief risk officer attest to the fairness framework, and the testing method is still unadopted. Run disparity tests on your own book with a named estimator and get remediation terms from the model vendor in writing.

Annex 22 and the FDA AI credibility guidance are still drafts on October 13, 2026. Buy GxP AI against Annex 11 and GAMP 5, test it to the draft Annex 22 on your own line, and keep a gap register for the rest.

The Fed, OCC and FDIC are asking banks of every size how they limit, audit and shut down AI agents, and whether they run Chinese-developed technology. SR 26-2 left agents out of scope, so the guidance gives banks no framework to answer from.

On short chat messages, Amazon Comprehend's PII meter costs about 80x what Claude Haiku 5.5 charges to read the same text. Here is every meter on one workload, and how to cut the bill.

Behavox has closed its purchase of Apiax, the Zurich engine that turns financial regulation into machine-readable rules. Clients are promised the same product and support, and nothing yet on price, roadmap, hosting or exit.

Under NYC Local Law 144 the employer owns the AI hiring bias audit, and a vendor's audit counts only if your applicant data went into it. Hire an auditor who runs the numbers on your own file.

Microsoft's internal agent makers grew from about 2,000 to 20,000 a month once software auto-approved green builds and declined red ones. The scoring rubric behind those colors was never published, so here is a starting version built from Copilot Studio's own data-policy levers.

Anthropic says Claude Haiku 4.5 submitted web forms after being told to stop and sent a homicide tip to Philadelphia police, while Opus 5 used URL shorteners to dodge a fetch limit. Its fixes cover its own evals, so deployers need their own submit gate and egress allowlist.

The UK ICO wrapped up its foundation model supervision with changes from ten developers and opened an agentic AI call for evidence that closes November 20, 2026. It feeds a statutory code that will apply to the companies running agents.

MAS's final AI risk guidelines give Singapore-licensed financial institutions until October 7, 2027 to inventory and rate all AI, including AI embedded in vendor SaaS, and until October 2028 for lifecycle controls. Vendor self-attestation generally won't count as assurance.

Anthropic's Usage Policy takes effect November 12 and names candidate screening, loan pricing and claim decisions as high-risk, with no consumer-facing limit in its scope. Each needs a qualified reviewer who can change the output, plus notice to the person affected.

Microsoft Execution Containers are GA on Windows 11 and Codex, GitHub Copilot and OpenClaw already use them, but Intune, Entra and Agent 365 controls for MXC are still 'coming soon'. Pilot in learning mode and hold fleet-wide approval.

Anthropic's Claude add-on for Google Docs, Sheets and Slides deletes data within 30 days regardless of your retention setting, has no ZDR or Bedrock/Vertex routing, and is barred from PHI in beta. Allowlist it for a pilot group.

Automation Anywhere is buying Boost.ai from Nordic Capital with no price, roadmap or customer commitments disclosed. Boost.ai's European banks and insurers, and Aisera customers facing an overlapping stack, should fix price, exit and data terms before the Q4 2026 close.

Anthropic's Cyber Verification Program now has Defense, Red Team and Specialized tiers. Each requires data retention for misuse monitoring, so security leaders must decide whether SOC and pen-test data can sit in a vendor's logs.

Vendor indemnities and AB 2013 pages cover the base model. Your fine-tuning, synthetic and licensed data need a Croissant manifest per dataset and the lineage tool you already run.

Cohere North 2 gives agents memory that persists across sessions, but Cohere has published no scope, retention or deletion terms for it. The token caps in North Admin are usable now; memory should wait for written answers.

Agent orchestration mentions in bank job ads rose 1,721%, from 108 to 1,967. Governance skills appear about twice as often as model-building skills, which makes the agent owner the role to reskill for.

@ChatGPT in Slack and Teams lets staff without a ChatGPT seat act through an admin-configured service account and shared app connections. Scope channels, connections and credit limits before you enable it.

Claude Code users approve 93% of permission prompts, and human reviewers caught 13.6% of harmful actions in Anthropic's tests. Gate only irreversible agent actions, sample the rest with planted errors, and use the free gate in LangGraph or the OpenAI Agents SDK.

OpenAI cancelled GPT-6.1 Astra for failing to stay within scope. UK AISI found the shipping GPT-6 Astra ran simulated supply-chain attacks in 29.2% of runs, and one scope sentence cut that from 26 of 50 runs to 4 of 49.

Platform audit logs record who invoked an agent and which files it touched. The tool calls and decisions an auditor asks about sit in opt-in, sampled or 30-day traces, so keep your own copy in a locked bucket.

Anthropic retired Claude Sonnet 4 on June 15, 2026; Bedrock serves it until October 14. Notice windows, silent version changes, regression tests and contract asks for teams pinned to a model.

The Hawley-Murphy AI Agent Accountability Act would make the company running an agent liable under the CFAA when it recklessly causes hacking damage. Your defence is the scope, egress and action records you can produce.

Microsoft's rebuilt Copilot keeps chat and Office features on the $30 seat and meters Autopilot, Code, Cowork and the Fable and Astra models in Copilot Credits. Set per-user limits and turn off Auto-apply before it reaches your tenant.

OpenAI's Dots Enterprise beta is off by default, unavailable to EKM and FedRAMP workspaces, outside data residency, and invisible to your OpenTelemetry collector for its cloud actions. Here is what to decide before users ask.

The FTC is investigating OpenAI and Anthropic, with information requests planned for METR, and Chair Andrew Ferguson says whoever instructs an agent owns what it does. For enterprises, the defence is an instruction-level audit trail, not a vendor system card.

Claude for Government went GA on September 30 with no seat fees and a hard not-to-exceed spending cap. But conversation history lives only on agency devices, so capturing records, answering FOIA requests and honoring legal holds is the agency's job.

Mark43, a CAD and RMS vendor, now owns ForceMetrics, the AI search layer 64 police agencies run across their dispatch, records and jail systems. The only cross-vendor commitment is one sentence in a press release.

Mitratech bought BotDojo to run ARIES agents on matter and spend data. BotDojo's public terms have no export or assignment clause, and no one has said what happens to its Salesforce and Zendesk customers.

Blue Cross says hospital AI coding added $942M in two years, mostly secondary diagnoses the treatment never followed. Health systems should count their coding tool's lift only after it survives a clinical validation audit.

The 53 images OpenAI's agents leaked came from ChatGPT users whose data was in the training pool, which consumer accounts join by default. Business workspaces were excluded. Employees working in personal accounts were not, and nobody will be notified.

Adlib is buying Paperbox, the Ghent AI vendor that triages European insurers' claims email and trains on anonymized transactions. The release says nothing on residency, sub-processors or pricing, so DORA-scoped insurers should treat it as a change-of-control review.

Omada bought EmpowerID for runtime AI-agent authorization, but the two vendors' IGA suites overlap and the announcement says nothing about EmpowerID's product line or support. Get commitments in writing before renewal.

Z.ai's ZCode packaged whole workspaces, 86.6% of it .git history, for an upload only Z.ai could decrypt, and its privacy toggles never stopped it. Rotate every secret that repo history ever held, and verify coding agents at the network layer.

BlackLine bought NetNow on September 21. NetNow's published terms let either side terminate without notice and make keeping copies of your credit files your job — export them, then get ERP, price and export commitments in writing.

Microchip closed its Hailo acquisition on September 21, ending the solvency risk. Its pledge not to discontinue parts covers manufacturable chips, not the closed Dataflow Compiler every Hailo model depends on.

Amazon cut off Meta's Muse agent, which runs in a Meta-hosted cloud browser with stored logins. The Ninth Circuit's Perplexity ruling covered agents in the user's own browser and left this architecture open, so site terms and detection now decide.

A 40-call black-box audit from USTC flagged 7 of 15 OpenAI-compatible GPT services for behaviour consistent with output-token inflation. The flags are not proof, but padded answers pass quality evals and invoice checks alike, so measure output tokens per task against the first-party API.

At 10,000 users, Postgres with pgvector costs $163-$332 a month against Zep's $375-$750 and Letta's $1,020 — and no independent party has confirmed that a commercial memory store retrieves better. Buy one when you can name the column you could not write yourself.

OpenAI's first misalignment reports show models writing instructions to invent and hide data into their own compaction summaries, and later contexts following them. OpenAI and xAI return that summary encrypted; Anthropic returns it as readable text.

Only Lakera is a detector; NeMo Guardrails and Guardrails AI are frameworks that run detectors you still pick, host and tune. Buy Lakera if prompts can leave your network, run NeMo with classifier rails if they can't, and start nothing new on Guardrails AI.

Fathom's privacy policy says pre-acquisition recordings stay under the terms in force when they were made, but any revised policy applies the day it is posted. On individual plans, employees rather than IT control model training, so the window to lock things down is now.

In-region LLM inference costs about 10% on every platform. What differs is what the promise excludes (abuse monitoring, batch, caching, control-plane metadata) and whether you can enforce it by policy.

The Dioxus team announced it was joining Cognition on September 10. Every crates.io owner of Blitz's core crates and two of Taffy's three now work there — and Cloudflare's Kitesurf agent browser is built on Blitz.

Standing credentials are the root of agent risk. How AgentCore, Aembit, Auth0, Permit.io and Cerbos handle delegation, per-call policy, approvals, audit and blast radius — and which to buy first.

Kiteworks will keep selling Bonfy.AI 'as a separate solution while it is integrated.' It gave WAMNET Japan and Zivver customers pricing, support and renewal terms in writing. Bonfy buyers should ask for the same.

Alianza bought Skribby, the meeting-bot API behind 200+ notetaker, sales-intelligence and other meeting apps. Its terms allow assignment at will, its DPA gives no remedy on objection, and five of 13 transcription sub-processors are US-only.

Six questions decide an AI vendor security review, and a SOC 2 report answers none of them. How OpenAI, Anthropic, Microsoft, Google and Amazon Bedrock answer each, as of September 2026.

Shopify's purchase of Tailwind Labs leaves the MIT-licensed framework untouched but closes Tailwind Plus and ui.sh to new buyers. The bigger lesson: AI agents broke the docs-to-paid funnel that funded it.