Adobe’s Digital Media CTO Ely Greenfield says a broad “AI-generated” label can obscure the difference between automating an edit and generating an image. In an interview with Digital Trends, he argues that provenance records such as C2PA’s Content Credentials can give people a more useful account of a file’s origin and changes, without settling whether it is true or who deserves credit.
Watch Desk analysis
What happened
Greenfield discussed C2PA, creative tools and Adobe’s approach to outside AI agents in an interview at Adobe MAX. C2PA is an open standard for attaching signed records of a file’s origin and edit history. Digital Trends notes that camera makers including Leica, Nikon and Canon have adopted it in some products and workflows.
That record has limits: it can show who supplied information about a file and what edits were recorded, but it cannot prove an image depicts the truth or measure how much human judgement went into making it. A file without credentials is not automatically fake; it may simply have no verifiable record attached. Read the Digital Trends interview.
The conversation also covered Adobe’s local MCP integrations, which the article says will let outside agents work with native files in Photoshop, Illustrator, Premiere and After Effects. What those agents can do depends on the integration and the access a user authorises. Greenfield also described Adobe’s own Creative Agent as drawing on context such as a creator’s style, brand and current work.
Why it matters
“AI used” can cover a phone processing a photograph, an agent applying a repeatable edit, or a model generating an image from scratch. Those are not the same creative act, and a single label risks making them sound interchangeable. Provenance can help show a file’s history, but it cannot do the harder work of deciding what that history means for originality or credit.
The MCP discussion puts a practical choice alongside that debate: creators may be able to bring outside agents into familiar applications, but the precise permissions and commercial terms matter. Openness is useful; “the agent can reach my project files” is not a detail to leave to vibes.
Our read
The strongest point here is the distinction between recording what happened to a file and judging the work itself. C2PA can offer a trail, not a verdict. That is a more useful promise than treating a badge as proof of authenticity, and a reminder that provenance only helps when people can understand what it records and what it does not.
What to watch
- Which creative apps and tasks Adobe’s local MCP integrations actually expose to outside agents.
- How users control access to project files and other creator data.
- Whether provenance tools can explain the difference between assisted editing and generated content without pretending to settle authorship.
Discussion spark: Should creative tools label every use of AI, or would a record of what the AI actually did give audiences more useful information?
Sources and evidence
- Source update (16 November 2026, 07:57 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.