Digimarc adds provenance, audit, and verification controls for AI agent workflows

Digimarc has announced new provenance and verification infrastructure designed to secure autonomous and AI-enabled workflows.

As enterprises increasingly adopt AI systems capable of generating content, orchestrating workflows, and taking action with minimal human intervention, establishing trusted provenance and verifiable authenticity is becoming mission critical. Digimarc’s new capabilities are designed to help organizations determine whether digital content and artifacts produced or consumed by autonomous AI agents can be trusted before downstream action occurs.

The OWASP Top 10 for Agentic Applications 2026 identifies artifact integrity, supply chain vulnerabilities, and audit non-repudiation among the highest-impact risks facing agentic deployments today. Without cryptographically verifiable records of what agents consumed and produced, and under what authority they acted, organizations cannot confidently audit agent behavior, satisfy emerging regulatory requirements, or defend against the tampering and manipulation scenarios the OWASP framework was created to defeat.

Grounded in the Coalition for Content Provenance and Authenticity (C2PA) standard, the open specification adopted by organizations including Adobe, Google, Microsoft, OpenAI, and others, Digimarc’s approach combines provenance, verification, and audit capabilities into a standards-based trust layer for AI-powered workflows.

Digimarc extends C2PA’s output attestation capability with a trust enforcement layer purpose-built for agentic environments. Every provenance seal is policy-gated, issued only when the identity of the agent, the integrity of the artifact, and the timing of the request all satisfy defined trust criteria.

The initial release centers on a new Model Context Protocol (MCP) server that enables systems and orchestration frameworks to stamp, verify, log, audit, and retrieve provenance information through MCP-compatible interfaces. By exposing provenance as a native capability within modern workflow architectures, organizations can introduce trusted verification and traceability directly into content and automation pipelines without requiring systems to manage the underlying cryptographic infrastructure themselves.

The MCP server is backed by Digimarc’s Illuminate platform and the company’s decades of experience in authentication, watermarking, and trusted identification technologies. For workflows requiring additional durability, Digimarc can combine C2PA manifests with its watermarking technology to help maintain provenance continuity even when metadata is altered or removed during processing.

“Organizations are deploying autonomous agents that produce, consume, and act on content at machine speed, and most have no reliable way to verify that the content those agents touched is genuine, unaltered, and attributable. Gartner projects that by 2028, 25% of enterprise AI applications will experience multiple security-related incidents annually,” said Ken Sickles, EVP and Chief Product Officer at Digimarc.

“We built this solution because provenance cannot be an afterthought bolted on after deployment. It has to be atomic with the agent’s work, enforced by the runtime, and verifiable by anyone downstream. By grounding our approach in open standards like C2PA and exposing these capabilities through our MCP server, we’re helping to make trusted provenance a native capability for modern digital workflows,” Sickles added.

Digimarc is launching the initiative with a select group of early build partners and platform collaborators to help shape real-world use cases and future roadmap priorities. Initial areas of focus include trusted content workflows, provenance verification, and governed traceability across emerging automation ecosystems.