Instagram will demote AI-generated influencers that don't self-label
Meta is tightening the rules for AI personas on Instagram, while OpenAI quietly tests a billing model that could reshape how enterprises buy AI.
Instagram will start limiting the reach of AI-generated persona accounts that fail to use its new "AI-generated profile" label, the company said in a blog post on Sunday. The policy replaces the older, optional "AI creator" tag with a mandatory designation for profiles that "feature an AI-generated person." Accounts that don't comply will be excluded from recommendations in Reels and Explore — the two surfaces that drive most discovery on the platform. Owners can appeal if they believe they were wrongly flagged. The move follows months of reporting from The Guardian and The New York Times documenting how AI-generated influencers have been used to promote health supplements and other products without disclosing that the face behind the account isn't human. Instagram isn't going after other kinds of AI-generated content — the rule is narrow, targeting specifically the persona accounts that blur the line between synthetic and real. That narrowness is the interesting part: Meta is drawing a line at identity, not at content, which suggests the company sees the reputational risk of fake humans as distinct from the broader flood of AI-generated posts.
OpenAI has begun letting some of its largest enterprise customers pay only when its AI actually completes a task, according to The Information. The arrangement is limited to select major accounts and hasn't been announced publicly — the terms, the customers, and the prices remain unknown. The industry calls this outcome-based pricing, and it's already the norm in customer support: Intercom charges $0.99 per conversation its Fin agent resolves, Zendesk restricts billing to "Verified Resolutions" confirmed by an LLM within 72 hours, and Salesforce has been working through the same question with its Flex Credits system. For OpenAI, the shift is significant because it means accepting the risk that the work doesn't complete — a meaningful transfer from buyer to vendor. The harder problem is defining success for the multi-step agentic work OpenAI has been pushing toward, where "completion" is a matter of judgment rather than a database field. A Futurum Group survey found 43% of AI buyers prefer consumption-based pricing and 27% prefer outcome-based, with fewer than one in five still wanting per-seat licenses. The fact that the largest model vendor is testing the model — even selectively — suggests the token-billing era for enterprise AI may be shorter than it looks.
Do you think outcome-based pricing will become the default for enterprise AI, or will token billing survive alongside it? Tell us in the comments.
