In January of this year, a Microsoft employee named Shane Jones wrote a letter to Washington State Attorney General Bob Ferguson and several U.S. senators and representatives. Jones claimed that he discovered a problem with the guardrail mechanism of Designer, Microsoft's artificial intelligence-based art production tool (formerly Bing's image creation tool). Today, Jones made a new claim, saying that Designer can create violent and sexually themed images through some text prompts, which is not allowed by Microsoft.
According to CNBC, Jones sent a new letter today to US Federal Trade Commission Chairman Lina Khan (Lina Khan), and also sent another letter to Microsoft's board of directors. In the letter to Khan, Jones claimed that he had urged Microsoft to remove Designer from public use for several months until new and better safeguards were put in place.
Jones claims that Microsoft refused to remove "Designer", so he is now asking the company to make new disclosures about what "Designer" can create and change the Android app's rating so that it is no longer rated E (for everyone).
Jones claimed that he entered "Pro-choice" (support choice) into Designer, and Designer then produced a large number of violent cartoon images. The images, viewed by CNBC, include a fanged demon preparing to eat a baby, Darth Vader holding a lightsaber next to a mutated baby, and a handheld drill-like device labeled "Support Choice" being used on a fully grown baby.
Jones also claims that "Designer" can be used to create Disney's copyrighted characters. The article said the images it saw included "Star Wars-branded Bud Light cans and Snow White's likeness on soda."
In a statement to CNBC, a Microsoft spokesperson said the company appreciates the efforts of its employees to test its technologies and services to help them use them more securely. The spokesperson added:
When it comes to security bypasses or concerns that may have potential impact on our services or partners, we have established robust internal reporting channels to properly investigate and remediate any issues, and we encourage employees to utilize these channels so that we can appropriately validate and test their concerns.
In February, Google shut down the AI image generator used in its Gemini AI chatbot after discovering it could be used to create racially offensive images. The company said it is setting up new safeguards so that the same images will not be generated again when the feature is restored.
Jones's new claims come as new research from the Center for Countering Digital Hate shows that artificial intelligence art makers, including from Microsoft, could easily be used to create images designed to create disinformation about elections and candidates.