Major AI labs commit to shared policy token for publisher training opt-outs

OpenAI, Google, Microsoft and Anthropic stated they will honour a new shared ‘policy token’ signal that publishers can place in robots.txt files to indicate content must not be used for AI training. The mechanism allows a single opt-out instruction to apply across multiple providers instead of separate per-company settings. Anthropic has already adopted the signal; OpenAI indicated implementation in coming months and Google and Microsoft the following year. The development creates a practical industry-wide control for content owners over training-data access.