OP-ED: Enter, the age of provenance — or the week AI started tagging itself

The age of provenance is upon us as AI models and platforms now tag and detect AI-generated content. This forces us to define authorship, disclosure, and the limits of human judgment. Companies like Anthropic, Substack, and LinkedIn are taking different approaches to address this issue. Anthropic plans to launch new Claude models in the EU on or after August 2, 2026, which will support machine-readable marking.

Substack has introduced "Scan for AI text," powered by Pangram, to estimate human-written or AI-assisted writing. Creators can also add a "How I make this" statement to explain their use of AI. LinkedIn is working on reducing the distribution of low-effort AI-generated content that lacks genuine perspective or substance. The underlying theme across these interventions is transparency and the importance of human judgment.

However, the question remains: what exactly counts as human work? As AI becomes more integrated into our daily lives and work processes, it becomes increasingly difficult to determine the balance between human effort and AI assistance. While AI can accelerate capability, it can also substitute for capabilities the author possesses.

This distinction is crucial to understanding the value created by AI-assisted work. Ultimately, the economics of time and effort play a significant role in the adoption of AI tools. If AI can help professionals complete tasks more efficiently, they will likely use it frequently. As AI becomes more prevalent, governance and regulation may lag behind, leading to a normalization of AI assistance in various domains.