Home · Technology · Nov 21 archive
OpenAI Launches Tools to Combat AI-Generated CSAM
Confirmed
In Short: OpenAI has released a new image classifier and in-chat warnings in ChatGPT to combat child sexual abuse material (CSAM).

On November 15, 2026, OpenAI announced a suite of tools aimed at detecting and preventing the creation and distribution of child sexual abuse material (CSAM) across AI platforms.
The centerpiece is an image classifier designed to detect CSAM, including images generated by AI. During early testing with partners, the classifier flagged 48% fewer images while identifying the same number of abusive images.
OpenAI is offering the classifier free through its API to vetted organizations, and it will be integrated into ChatGPT to show alerts to users attempting to access or generate such content.
Samuel Wang, Member of Technical Staff at SafetyKit, praised the classifier's performance, noting it catches unknown CSAM that hash matching misses, with false positives below 1% on legal adult content.
OpenAI is also providing guidance for API developers on integrating perceptual hash matching alongside classifiers for novel content, and working with the Tech Coalition and ROOST to share access.
OpenAI's efforts include a specific version of ChatGPT for teen users aged 13–17, launched in August 2026, with stronger safety protections and learning-focused features.
OpenAI's new framework for publicly disclosing AI misalignment incidents aims to set standards across the industry.
In a separate development, OpenAI announced that its swarm of 10,000 AI agents had solved the Navier-Stokes problem, one of the seven Millennium Prize problems, in less than four days.
At the United Nations, OpenAI's CEO Sam Altman and Anthropic's CEO Dario Amodei have urged for international coordination on AI development.
OpenAI's GPT-6 Astra, introduced on April 9, 2026, is being used to develop these new tools.
The FTC is investigating OpenAI, Anthropic, and METR over consumer risks from autonomous AI, with METR also a target.
OpenAI is pushing deeper into AI-enabled commerce, introducing dots agents during its annual developer conference in late September.
What this adds
OpenAI's new tools are designed to combat the strain on older safety tools due to AI-generated abuse material.
OpenAI's guidance for developers includes integrating perceptual hash matching alongside classifiers for novel content.
OpenAI's efforts are part of a broader initiative to strengthen online child protection and combat CSAM.
Background
OpenAI announced a suite of tools aimed at combating child sexual abuse material (CSAM) on November 15, 2026.
What's confirmed
- The company is pairing detection tech, in-chat warnings, and developer guidance as AI-generated abuse material strains older safety tools Mon 16 Nov 2026 · 12:03 EST · 3 min read OpenAI has announced a new set of tools aimed at child sexual abuse material (CSAM).
- On November 15, 2026, OpenAI published CSAM guidance for API developers that walks through integrating perceptual hash matching alongside classifiers for novel content.
- During early testing with partners, OpenAI found that the classifier flagged 48% fewer images while identifying the same number of abusive images.
- OpenAI has launched a free AI-powered tool to help organisations detect child sexual abuse material (CSAM), including AI-generated images, as part of a broader effort to strengthen online child protection.
- OpenAI is working with the Tech Coalition and ROOST to share access.
- "Our goal here is to really focus on merchants and our network partners and start building out a really great foundation that's safe and easy to use," Chaya Nayak, who leads product for monetization and payments at OpenAI, told CNBC in an interview.
- With the initial launch, OpenAI is working to build relationships with merchants and position itself as a serious platform for shoppers, especially as vendors are beginning to experiment with personal agents.
- OpenAI has pushed deeper into AI-enabled commerce over the last year, but some of its early attempts failed to take off.
- OpenAI introduced the dots agents during its annual developer conference in late September, shortly after Meta's Muse offering topped the App Store charts.
- OpenAI said its dots agents will never actually access a user's card details.
- It assigns a score to an image, but OpenAI says that score alone does not confirm abuse, establish its legal status, or determine whether a report is required.
- The warning explains the potential harm, directs users to specialist support, and notes that AI-generated material can also be harmful OpenAI will remove the flagged conversation and pause the user’s access to ChatGPT for 24 hours.
What's still developing
- We keep to what Reuters and OpenAI have published.
- He pointed to OpenAI's reliance on what it calls "iterative deployment", where systems are released and protections strengthened as problems appear.
- METR, a Berkeley, California research group that OpenAI and Anthropic have both used to run independent reviews of their agentic AI, is also a target, the official said.
- An OpenAI spokesperson described attackers adapting their tactics as the company introduced new defenses.
- In an evaluation involving user-uploaded images, OpenAI reported that the new classifier: Early-access partners also reported results from their evaluations: OpenAI has also published a CSAM classifier model card and an intake form for organisations interested in accessing the tool.
- According to OpenAI, the initiative aims to help businesses connect with ChatGPT's 1.2 billion weekly users as they explore products, compare options and make purchasing decisions.
- According to OpenAI, the findings were particularly significant because several earlier AI models had already examined the browser, yet these security weaknesses had remained undetected.
- The company said the models were developed using methods similar to those behind GPT-6 Astra, which OpenAI introduced earlier this month.
- According to OpenAI, the system uses a library of native components and a compiler that progressively builds the interface as the response is generated.
- Of the two new models, OpenAI says that Sol is “smarter, more efficient, and lets you complete more tasks on the same paid plan” as a “go-to model for everyday work.” Astra, meanwhile, is tuned for the highest-quality work in professional use cases such as documents, slides, coding, and more: GPT‑6.2 Astra sets a new quality bar for AI in complex professional work.
- But, now, OpenAI is stepping up to the challenge.
- Together, they cover the user, the builder, and the investigator.
Sources
- 9To5Googlelink
- Crypto Briefinglink
- Innovation Villagelink
- International Business Times, Singapore Editionlink
- Lawnextlink
- MobiGyaanlink
- SQ Magazinelink
- TechCrunchlink
- TechTimes Vietnamlink
- Techiexpertlink
- The Indian Expresslink
- WIREDlink
- Winbuzzerlink
- news.lavx.hulink
- shattered.iolink
- Shatteredlink
- AI Master — video link
