(Credit: Shutterstock / Mojahid Mottakin)
A Microsoft AI engineer sent letters to the US Federal Trade Commission and Microsoft's board Wednesday over concerns that Microsoft's Copilot Designer is generating "harmful content."
Copilot Designer, which is powered by OpenAI's DALL-E 3, is creating offensive images containing "political bias, underaged drinking and drug use, misuse of corporate trademarks and copyrights, conspiracy theories, and religion," writes Microsoft Principal Software Engineering Manager Shane Jones, who has been at the tech giant for over six years. CNBC reports that it replicated similar images with the tool this week.
Jones says Copilot Designer, also known as the Bing Create tool, will often add "sexually objectified" women in its compositions without being asked. The engineer is concerned by the disturbing content present in the AI tool's outputs as well as the fact that Microsoft doesn't currently warn users about the AI's potentially offensive or problematic content.
The image-generating tool's website doesn't clearly list any warnings at time of writing, but an FAQ section on its site states: "We have put controls in place to prevent the generation of harmful images. When our system detects that a potentially harmful image could be generated by a prompt, it automatically blocks the prompt and informs the user."
Jones claims that he has repeatedly tried to get Microsoft to add an age rating or restrictions to Copilot Designer, but Microsoft rejected the request.
"Over the last three months, I have repeatedly urged Microsoft to remove Copilot Designer from public use until better safeguards could be put in place," he writes. "Having refused that recommendation, I have suggested they at least add disclosures to the product and change the rating on their Android app from 'E for Everyone' to 'Mature 17+.' Again, they have failed to implement these changes and continue to market the product to 'Anyone. Anywhere. Any device.'"
The engineer wants the FTC to inform the public about Copilot Designer's inherent problems and get Microsoft to properly disclose the types of content Copilot Designer can produce. The FTC confirmed to PCMag via email Wednesday that it received Jones's letter, but did not have any additional comment.
Jones's latest letter follows one sent in January to four US Senators expressing similar concerns. Jones also previously published a letter to OpenAI on his LinkedIn regarding the issue but says Microsoft, which owns LinkedIn, persuaded him to delete the December post.
In Jones's letter to OpenAI, the engineer said DALL-E 3 poses a large public risk "to the mental health of some of our most vulnerable populations including children and those impacted by violence including mass shootings, domestic violence, and hate crimes."
Microsoft previously said it was investigating the issues and shared that OpenAI also investigated the engineer's reports. Jones said Microsoft closed one of his reports without taking action. But OpenAI heard him out, at least. "They took the time to meet with me and were very grateful for the information I shared about the vulnerability in their products," Jones tells PCMag in a message.
In a statement, a Microsoft spokesperson said the company is "committed to addressing any and all concerns employees have in accordance with our company policies and appreciate the employee’s effort in studying and testing our latest technology to further enhance its safety.
"When it comes to safety bypasses or concerns that could have a potential impact on our services or our partners, we have established in-product user feedback tools and robust internal reporting channels to properly investigate, prioritize and remediate any issues, which we recommended that the employee utilize so we could appropriately validate and test his concerns," the spokesperson adds. "We have also facilitated meetings with product leadership and our Office of Responsible AI to review these reports and are continuously incorporating this feedback to strengthen our existing safety systems to provide a safe and positive experience for everyone."
Copilot Designer isn't the only AI image tool that's caused controversy, however. Google paused its Gemini AI from being able to create human images last month after the AI generated depictions of people of color in inaccurate historical representations.
PCMag has reached out to OpenAI for comment.
Editors' Note: This story was updated with comment from Microsoft.