Ex-Google Trio Launches Open Creator Graph, a Consent Layer Between Creators and AI
Singapore-based Sparkonomy launched the Open Creator Graph today. The free platform lets creators declare how AI systems may use their name, image, voice, and likeness. Creators claim a record by signing in with an Instagram, YouTube, TikTok, or X account, then set consent terms for each part of their identity. The record publishes in human-readable form and in machine-readable formats: JSON-LD, XML, TDMRep, and plaintext. Search engines and AI systems can find and honor the terms.
The platform launches with one million pre-populated creator records available to claim.
Behind the launch are three former Google executives. Guneet Singh led YouTube’s growth across APAC and founded Google’s Ignition Labs. Megha Thareja Tyagi held director-level Google roles across India and Southeast Asia, with earlier stints at PayPal and American Express. Vipasha Joshi launched Google+ in India and later ran Jellysmack’s India business as country manager.
The study behind the launch
Sparkonomy timed the launch to its own research, the AI Misrepresentation Study, or AMISS. The company submitted 50,000 queries across 10,000 creators in September 2026. It asked ChatGPT, Gemini, Grok, Muse, and Qwen to describe each creator from existing knowledge. One creator in 10,000 was described accurately by all five engines. Sixty-seven percent received no accurate description from any engine.
Even among creators with more than five million followers, only 37% got an accurate description anywhere.
Those numbers warrant caution. The study is Sparkonomy’s own, and the release does not define what counts as an “accurate description.” Independent measurements paint a less dire picture of fact accuracy.
An Entidex study of 50 creators and shows found engines agreed with a verified record on about 72% of facts. A Searchable review of 32,000 brand facts found 9.2% false. Different questions, different yardsticks. The misrepresentation problem is real, but AMISS measures it by the strictest possible standard.
How the record works
Each record carries consent terms for six identity attributes: name, image, likeness, voice, movement, and signature. Creators pick one of three levels for each. “Allowed” permits any AI use. “Conditional” lets AI represent the creator as they are, while deepfakes and voice clones require permission first. “Prohibited” blocks display, training, and generation entirely.
Two design choices stand out. Where a creator has not set terms, the record reads “unknown” — explicitly not permission. Every issued consent certificate is digitally signed and added to a tamper-evident public log. A qualified service under the EU’s eIDAS framework timestamps it. That makes a creator’s declaration verifiable evidence in a dispute.
The legal tailwind
The launch lands amid genuine regulatory movement. EU copyright law already requires rights reservations on publicly available content to be machine-readable. The EU AI Act obliges general-purpose AI providers to identify and comply with them.
In the United States, Tennessee’s ELVIS Act has protected voice and likeness against unauthorized digital replicas since July 2024. The federal NO FAKES Act advanced unanimously out of the Senate Judiciary Committee in June 2026. It now sits on the Senate calendar — still a bill, not a law. Indian courts have granted celebrities protection over name, voice, and likeness against AI imitation.
Co-founder Megha Thareja Tyagi pointed to India. Officials there are exploring a model that would channel royalties to creators when AI commercializes their work. “Every creator who puts themselves online should be able to protect what their work earns today, and share in the value this AI era is building,” she said.
The question that decides everything
The platform’s fate rests on a single variable: whether AI companies read and honor the signals. Nothing compels them to. Sparkonomy’s co-founder Guneet Singh argues the multiple publication formats leave “no technical excuse.” Technical availability, though, is not the same as adoption. Precedent cuts both ways.
Spawning AI’s “Do Not Train” registry has operated since 2023, and TDM·AI maintains federated registries of training preferences. Opt-out compliance across the industry remains uneven.
The “first” claim also deserves scrutiny. Sparkonomy calls the Open Creator Graph “the first free, public place to bring a creator’s identity and AI consent together.” Content-level opt-out registries predate it; the novelty is combining identity verification with consent in one record. The distinction matters, but so does the phrase’s precision.
One more thing is worth knowing. Sparkonomy describes the Graph as “the first step,” to be followed by commercial AI solutions for the creator economy. The record itself is free, with no subscription and no follower minimum. The business that will eventually monetize around it does not exist yet, on the company’s own telling.
That makes the Open Creator Graph either critical public infrastructure or a very good sales funnel. It is too early to say which.
Editor’s Note
This article is based on a press release from Sparkonomy distributed on September 29, 2026. TechRecast independently verified the founders’ backgrounds, the platform’s launch coverage, the status of the NO FAKES Act and the ELVIS Act, and the existence of prior consent registries. The AMISS study is Sparkonomy’s own research; its methodology and accuracy criteria are not published in the release, and its findings are reported here with that caveat. Independent studies with different methods measure the misrepresentation problem less severely.