Whistleblower says OpenAI’s DALL-E should be closed to the public.
Insider Silenced
Things are heating up for Microsoft.
First, one of the company’s AI programs was blamed for the deepfake pornographic images of Taylor Swift flooding social media, which were so bad that even the White House chimed in with a statement.
Now, Shane Jones, a Microsoft software engineer, has come forward with claims that he had found security holes in OpenAI’s DALL-E that could be harnessed to make abusive images like the ones targeting Swift, according to GeekWire, but that he was told to clam up about the issue.
Jones said he found the security issue in early December and reported it to Microsoft, which instructed him to inform OpenAI, which has a close partnership with Microsoft.
After he let OpenAI know about the issue, he never heard back, the news outlet reports. He then went public about the security flaw in a LinkedIn post while also urging OpenAI to shutter DALL-E because he felt the program posed a public safety risk.
After he told Microsoft about the post, Jones says, Microsoft’s legal team ordered him to delete it.
Whistle Blower
Unsatisfied by OpenAI’s and Microsoft’s silence and inaction, Jones sent a letter to Washington state’s representatives in Congress and the state Attorney General Bob Ferguson, warning about the security issues in AI.
In the letter, Jones cited the Swift images, which were created on Microsoft’s Designer, built partly on DALL-E technology.
“The vulnerabilities in DALL·E 3, and products like Microsoft Designer that use DALL·E 3, makes it easier for people to abuse AI in generating harmful images. Microsoft was aware of these vulnerabilities and the potential for abuse,” he wrote.
OpenAI issued a statement to GeekWire that said the security issue Jones had flagged “does not bypass our safety systems.”
In addition, Microsoft had said any security bypasses that led to the horrible Swift deepfakes have been shuttered.
Since these AI tools have become widely available, hackers have often jailbroken them and bypassed their security features, including on OpenAI products — which shows how incredibly vulnerable this tech still is to wrongdoing.
More on Taylor Swift: Microsoft CEO Horrified by AI-generated Taylor Swift Images
Share This Article