Microsoft’s AI imaging tool, powered by technology from Sam Altman’s OpenAI, has raised concerns among employees at the tech giant due to its potential to generate sexually explicit and violent images that may pose risks to the public.
Shane Jones, a principal software engineering manager with a six-year tenure at Microsoft, recently took a bold step by addressing these concerns in letters to the company’s board of directors and the Federal Trade Commission. In his communication, Jones highlighted apprehensions regarding Microsoft’s handling of ethical considerations in AI development.
Jones, known for his role as a “red teamer” responsible for proactively testing AI technologies for vulnerabilities, specifically pointed out issues related to Microsoft’s Copilot Designer. This AI image generator, akin to OpenAI’s DALL-E text-to-image model, has come under scrutiny for producing problematic content during Jones’ testing.
During his evaluation of Copilot Designer over the past three months, Jones discovered troubling depictions ranging from demons, monsters, and contentious social issues like abortion rights to concerning imagery of teenagers with firearms, sexualized portrayals of women in violent scenarios, and instances of underage substance use.
Despite bringing these issues to the attention of Microsoft’s leadership in December, Jones faced resistance in his efforts to address the flaws in Copilot Designer. The company redirected him to OpenAI, the entity behind technologies like ChatGPT and the underlying DALL-E 3 model powering Copilot Designer.
Jones’s attempts to engage OpenAI in rectifying the identified issues were met with silence, prompting him to share his concerns publicly on LinkedIn. However, Microsoft’s legal team intervened, instructing Jones to remove the post. Undeterred, Jones escalated his worries to the US Senate, advocating for stricter oversight of AI technologies like Copilot Designer.
In a letter addressed to FTC Chair Lina Khan, Jones reiterated his plea for Microsoft to temporarily halt the deployment of Copilot Designer until comprehensive safeguards are in place. Despite his persistent advocacy for corrective actions, Microsoft continues to promote the product unabated, raising questions about the prioritization of ethical considerations in AI innovation.
