EU Pushes High-Risk AI Compliance to 2027, Disseqt Raises $8M for Agent Controls
The EU delayed Annex III high-risk obligations to December 2027, but enterprises still face near-term inventory, classification, and supplier-contract work. An $8 million round for agent-assurance vendor Disseqt signals runtime enforcement is becoming a distinct budget line.
EU moves high-risk AI deadlines, but near-term work remains
The EU's Digital Omnibus on AI postponed key high-risk system obligations to December 2, 2027 for Annex III systems—including recruitment and career-management tools—and to August 2, 2028 for high-risk AI embedded in regulated products. The regulation entered into force July 27, 2026.
The delay does not eliminate work in 2026. Prohibitions, AI-literacy requirements, general-purpose AI obligations, and Article 50 transparency requirements retain their original timelines, with some transparency deadlines starting December 2, 2026. Enterprises buying AI recruiting tools in 2027 will need to verify conformity assessments, CE marking, EU declarations of conformity, registration status, logging capabilities, incident reporting, and human-override functions before those systems go live.
The revised dates shift implementation spending from immediate certification toward AI inventories, use-case classification, supplier contract reviews, human-oversight procedures, and evidence collection in 2026 and 2027. Systems deployed before the new high-risk dates become subject to those obligations if they undergo a significant change in design or intended purpose, which means procurement teams need vendors to define and document what constitutes a material update.
What this means for enterprise buyers
The timeline change gives software vendors more runway to build conformity-assessment workflows, documentation systems, monitoring capabilities, and EU database-registration processes. Buyers should still compare vendors on readiness for ISO/IEC 42001, NIST AI RMF, privacy controls, cybersecurity posture, and sector-specific requirements rather than accepting a generic "EU AI Act compliant" claim.
Enterprises can defer some formal high-risk conformity work, but deferring AI inventories, use-case classification, and supplier reviews creates downstream risk. The Act applies obligations based on system characteristics and use cases, not vendor claims. A recruiting tool classified as high-risk in 2027 requires the same conformity evidence whether the vendor prepared early or at the deadline.
The change affects procurement schedules, contract language, and product-roadmap requirements across HR technology, regulated products, and other high-risk categories. Contracts signed in 2026 should include obligations for vendors to deliver conformity evidence, maintain registration, and report incidents on the new timeline.
Disseqt raises $8 million for agent runtime controls
Irish AI-assurance company Disseqt AI raised $8 million on November 18, 2026 to build an enterprise governance layer for AI agents and workflows. The company says its platform tests AI applications before production deployment, enforces organizational policies during operation, monitors agent behavior continuously, blocks policy-violating outputs, and preserves audit evidence.
Disseqt enters a market that includes AI observability vendors, model-risk management platforms, application-security tools, data-loss-prevention systems, and dedicated AI-governance software. Its stated differentiation is combining pre-deployment testing, runtime enforcement, and audit evidence for autonomous agents rather than offering a governance registry or documentation repository.
The round is an early signal that agent governance is becoming a distinct enterprise software category rather than an extension of existing GRC, security, or observability tools. Enterprises evaluating whether to handle agent controls through existing platforms or a dedicated control plane should demand measurable evidence: policy-violation detection rates, false-positive rates, latency added by runtime blocking, supported model and agent frameworks, retention periods, and exportable audit records.
The available announcement does not provide performance benchmarks, customer counts, pricing, or named deployments. Agent deployments may require a new recurring software budget for runtime controls in addition to model inference, security, and compliance spending. The absence of disclosed pricing and benchmark data makes Disseqt's commercial maturity difficult to compare with established governance and security platforms.
What to watch
The EU deadline changes should drive immediate enterprise planning even though some high-risk obligations moved to 2027 and 2028. Organizations waiting until 2027 to start AI inventories, use-case classification, and supplier contract reviews will compress implementation work into a narrower window.
The Disseqt round shows where the vendor market is moving: from static AI inventories and policy documents toward continuous testing, runtime intervention, and audit trails for autonomous systems. Enterprises should evaluate whether their existing security, observability, and GRC platforms can handle agent-specific controls or whether a dedicated layer is required. The answer depends on agent deployment scale, risk tolerance, regulatory obligations, and the ability of current tools to enforce policies at runtime rather than only logging violations after the fact.
Treat the Disseqt funding as a vendor-market signal, not proof of production-scale adoption. The growth of agent-assurance vendors indicates buyer demand for runtime enforcement and audit evidence, but enterprises need vendor performance data, pricing transparency, and customer references to evaluate commercial readiness.
Technology decisions, clearly explained.
Weekly analysis of the tools, platforms, and strategies that matter to B2B technology buyers. No fluff, no vendor spin.
