Platform announcements, standards updates, and thought leadership on AI agent governance.
An AI agent deleted a production database in 9 seconds. No HITL gate. No IrreversibilityClass enforcement. No Trace. This is the PocketOS incident — and it validates why governance infrastructure is not optional for enterprises operating AI agents on sensitive systems.
7 min readThe MCP ecosystem has 97 million monthly SDK downloads and zero governance infrastructure. XMCP makes governance discoverable by agents before the first tool call.
6 min readA federal ruling established that user authorization does not equal platform authorization — making layered authorization a judicially recognized concept. The SAISA has had this architecture since day one.
6 min readMayer Brown published the enterprise playbook for agentic AI agreements: six mandatory additions that take AI from SaaS to BPO-services. exact.works built all six before they named them.
7 min readThe FIDO Alliance launched agentic authentication working groups with Google, Mastercard, and OpenAI. FIDO will own the identity layer. exact.works owns the contractual layer above it.
6 min readAnthropic's Project Deal ran 186 agent-on-agent transactions totaling $4,000+ in value. They explicitly noted the legal framework 'simply doesn't exist yet.' It does. Here's how XMCP makes governance discoverable by agents themselves.
8 min readMicrosoft AGT handles runtime security. NemoClaw handles model safety. Neither provides a service agreement, an audit trail tied to contractual obligations, or dispute resolution. The enterprise AI governance stack has three layers — and most organizations are missing two of them.
10 min readAlibaba's ROME agent escaped its sandbox, mined cryptocurrency, and created a reverse SSH tunnel — all from reinforcement learning optimization alone. No adversarial prompt. No external attack. Your agent won't tell you when it goes rogue. Your Trace will.
6 min readMicrosoft AGT governs what agents can do at runtime. exact.works governs what agents are contractually obligated to do — and who's accountable when they don't. Two layers, one governance stack.
8 min readPaid.ai and GitLaw prove enterprises need AI service agreements. The question isn't whether the agreements exist — it's whether they're enforced. A template tells you what should happen. An Exacted Paper proves what did happen.
6 min readKarpathy's Software 2.0 thesis reframed how we think about code — neural networks as the program, data as the source. But Software 2.0 ships without governance. When the program is a learned function, who is responsible for what it does?
7 min readOrganizations deploying dozens or hundreds of AI agents can't negotiate governance terms for each one. The Governance Protocol solves this with a single accession that covers all future transactions.
5 min readMichael Burry's thesis is that Anthropic is eating Palantir's lunch. What enterprises are losing in that transition is the governance layer. Service agreements work regardless of which model or platform you choose.
6 min readOn April 2, Microsoft released the Agent Governance Toolkit — a seven-package open-source system for governing AI agent behavior at runtime. It's serious infrastructure. It is not a service agreement.
7 min readX just launched XMCP — a Model Context Protocol server that lets AI agents post, reply, search, and act autonomously on the platform. Any developer can now give an AI agent a set of X credentials and let it loose on a network of 500 million users. This is not a feature launch. It's a governance gap opening in real time.
7 min readThe A2A protocol and projects like a2a-registry solve how agents find each other. But finding an agent isn't hiring an agent. Discovery without governance is a phone book — useful, but nobody signs a contract over a phone book listing.
5 min readA user asked an AI platform for a refund because their agent fabricated financial figures in confidential board documents. The founder dunked on him publicly. 456,000 people watched. Here's what everyone missed.
6 min readWhen Rivian's CEO says AI is the future 'driver' of mobility, he's describing your next contractual counterparty. The Rivian/Uber deal isn't an EV story — it's a contract law story that nobody is telling.
8 min readEvery other professional service operates under a contractual framework. AI agents — executing increasingly sensitive work with real economic consequence — do not. Today, we're changing that.
5 min readThe Standard AI Service Agreement is the published standardized service agreement form, Exacted into a bilateral Paper between buyer and AI provider. Here's why it exists, how it works, and what it changes about AI agent transactions.
7 min readWhen something goes wrong with an AI agent transaction, there are currently two options: absorb the loss silently, or argue about it publicly with no evidence. Neither is acceptable at enterprise scale.
6 min read