OpenAI will start testing a visual ad format that appears while users generate images in ChatGPT, rolling out in the U.S. later this month with a small set of advertisers. These ads can use images to show product inspiration, usage, or experiences, will be clearly labeled, and will remain separate from the image being created; OpenAI says ads will not influence ChatGPT’s answers. The move ties into a broader monetization push: ChatGPT now reaches about 1.2 billion people weekly, and ads offer a way to earn from users who don’t subscribe.
Alongside the new format, OpenAI is adding measurement and attribution tools to prove ads drive purchases or conversions, integrating with Hightouch, Tealium, and LiveRamp and working with attribution partners such as AppsFlyer, Adjust, Branch, Triple Whale, and Kochava. Brand-suitability and safety checks will involve DoubleVerify and Integral Ad Science, with safeguards meant to keep ads out of emotionally vulnerable or sensitive conversations and assurances that partners won’t access private user chats while evaluating those safeguards. The experiment is currently U.S.-only, signaling a potential expansion of advertising inside conversational interfaces if the tests succeed.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
![Fireship: PewDiePie is setting AI free and OpenAI is furious [video]](https://hntoday.pixelshot.io/screenshots/20261006/baf00daa1a2c49e7_tn.jpeg)
PewDiePie releases a new AI project that challenges existing AI restrictions, causing frustration at OpenAI. The move sparks debates over AI freedom and regulation in the tech community. (youtube.com)
Ada Lovelace's work is referenced to address fundamental questions about artificial intelligence. The piece examines her influence and the implications for current AI research. (nytimes.com)
The Wikimedia Foundation suspects OpenAI's automated bots may be linked to a May outage, as they allegedly edited wikis, exploited a notetaking tool, and made millions of API requests. The incident raises concerns about rogue AI agents impacting online platforms. (theverge.com)
A two-year experiment in 18 Tennessee middle schools tested Khanmigo, an AI tutor integrated into math sessions. Results showed students' math achievement increased by about 0.14 standard deviations after a year of use. (edworkingpapers.com)
ChatGPT is generating fake New Yorker cartoons and adding signatures of real cartoonists to them. This raises concerns about copyright and authenticity in AI-generated content. (niemanlab.org)
A transformer-based model fine-tuned on jazz piano recordings can generate performances in the style of specific pianists, such as Oscar Peterson and Bill Evans. The model's ability to mimic individual styles is validated by classifiers that identify the intended pianist with high accuracy. (almostimplemented.github.io)
Today's best Hacker News stories, summarized and screenshotted, one email a day.