A Microsoft AI engineer claims that Copilot Designer Produces “Harmful Images.”
According to Shane Jones, an AI engineer at Microsoft, the Copilot Designer AI tool is producing graphics that glorify “the worst of humanity.”
A Microsoft AI developer expressed concerns that the company’s Copilot Designer is producing “harmful content” in letters to the US Federal Trade Commission and Microsoft’s board on Wednesday.
Based on OpenAI’s DALL-E 3, Copilot Designer generates objectionable visuals with “political bias, underaged drinking and drug use, misuse of corporate trademarks and copyrights, conspiracy theories, and religion,” according to Shane Jones, Principal Software Engineering Manager at Microsoft, who has worked for the company for more than six years. Similar photos were recreated this week using the program, according to CNBC.
Jones claims that without permission, Copilot Designer—also referred to as the Bing Create tool—will frequently include “sexually objectified” women in its compositions. The unsettling content in the AI tool’s outputs worries the developer, as does Microsoft’s current failure to alert users to potentially offensive or harmful stuff from the AI.
As of the time of writing, the image-generating tool’s website does not explicitly contain any cautions; however, it does state in its FAQ section that “We have put controls in place to prevent the generation of harmful images.” Our technology automatically suppresses prompts and notifies users when it determines that they may generate potentially hazardous images.”
According to Jones, he has made multiple attempts to persuade Microsoft to put limitations or an age rating to Copilot Designer, but each time, Microsoft has turned him down.
“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,” the author says. “As they rejected that advice, I’ve proposed that they at the very least make disclosures about the product and update the Android app’s rating from ‘E for Everyone’ to ‘Mature 17+.’ Once more, they have not succeeded in putting these adjustments into practice and are still marketing the product to “Anyone. Anywhere. Any device.”
The engineer wants Microsoft to appropriately reveal the kinds of content that Copilot Designer is capable of producing, and he wants the FTC to educate the public about the inherent issues with Copilot Designer. The FTC did not have any further comments, but it did confirm via email on Wednesday that it received Jones’s letter to PCMag.
Jones’s most recent letter is the follow-up to one that was written to four US senators in January, voicing identical concerns. Jones claims that Microsoft, the company that owns LinkedIn, convinced him to remove the letter he had previously published on his profile in December to OpenAI about the matter.
DALL-E 3 poses a significant risk to the mental health of some of our most vulnerable groups, including children and people affected by violence, such as mass shootings, domestic abuse, and hate crimes, according to Jones, the engineer, in a letter to OpenAI.
In a prior statement, Microsoft stated that it was looking into the problems and revealed that OpenAI looked into the reports from the engineers. According to Jones, Microsoft closed one of his reports without following through. At least OpenAI gave him a chance to speak.
Microsoft 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,” a representative for the firm said in a statement.
“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 continues. “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.”
Still, there are other AI image tools that have generated criticism besides Copilot Designer. After the Gemini AI produced erroneous historical renderings of people of color, Google suspended the AI’s ability to make human photos last month.