The White House announced voluntary commitments from leading AI firms, with the companies agree to combat AI-generated deepfakes.

Deepfakes have been a concern surrounding AI long before OpenAI made the technology accessible to everyday users. As AI-powered image generators have continued growing in popularity, there have been the inevitable abuse of such technologies.

The White House has been working with AI firms to try to establish safeguards designed to protect people from deepfakes, especially in the context of sexual abuse. According to the administration, Adobe, Anthropic, Cohere, Common Crawl, Microsoft, and OpenAI have all made varying commitments.

• Adobe, Anthropic, Cohere, Common Crawl, Microsoft, and OpenAI commit to responsibly sourcing their datasets and safeguarding them from image-based sexual abuse.
• Adobe, Anthropic, Cohere, Microsoft, and OpenAI commit to incorporating feedback loops and iterative stress-testing strategies in their development processes, to guard against AI models outputting image-based sexual abuse.
• Adobe, Anthropic, Cohere, Microsoft, and OpenAI, when appropriate and depending on the purpose of the model, commit to removing nude images from AI training datasets.

The White House also highlighted additional measures various companies have taken to combat the problem.

• Cash App and Square are curbing payment services for companies producing, soliciting, or publishing image-based sexual abuse, including through additional investments into resources, systems, and partnerships to detect and mitigate payments for image-based sexual abuse.
• Cash App and Square commit to expanding participation in industry groups and initiatives that support signal sharing to detect sextortion and other forms of known image-based sexual abuse to help detection and limit payment services.
• Google continues to take actions across its platforms to address image-based sexual abuse, including updates in July to its search engine to further combat non-consensual intimate images.
• GitHub, a Microsoft company, has updated its policies to prohibit the sharing of software tools that are designed for, encourage, promote, support, or suggest in any way the use of synthetic or manipulated media for the creation of non-consensual intimate imagery.
• Microsoft is partnering with StopNCII.org to pilot efforts to detect and delist duplicates of survivor-reported non-consensual intimate imagery in Bing’s search results; developing new public service announcements to promote trusted, authoritative resources about image-based sexual abuse for victims and survivors; and continuing to demote low quality content across its search engine.
• Meta continues to prohibit the promotion of applications or services to generate image-based sexual abuse on its platforms, has incorporated solutions like StopNCII and TakeItDown directly into its reporting systems, and announced it had removed around 63,000 Instagram accounts that were attempting to engage in financial sextortion scams in July. Meta also recently expanded its existing partnership with the Tech Coalition to include sharing signals about sextortion activity via the Lantern program, helping to disrupt this criminal activity across the wider internet.
• Snap Inc. commits to strengthening reporting processes and promoting resources for survivors of image-based sexual abuse through in-app tools and via their websites.

The announcement shows the effort companies and lawmakers are putting forth to ensure the safe deployment of AI, along with the challenges that are involved in accomplishing that.