Jones has previously warned that the platform can be easily tricked into creating inappropriate images from innocuous prompts. CNBC was able to recreate many of the scenarios Jones highlighted using the standard version of the software. In response to Jones' initial complaint in January, OpenAI stated that the prompting technique he shared does not bypass security systems and that they have developed robust image classifiers to steer the model away from generating harmful images. Microsoft also stated that they have established robust internal reporting channels to investigate and remediate any issues. However, Jones' concerns appear to remain unaddressed.
Key takeaways:
- Microsoft engineer Shane Jones has raised concerns about the safety of OpenAI’s DALL-E 3, suggesting it has security vulnerabilities that allow it to create violent or sexually explicit images.
- Jones has taken his complaint to the FTC, urging Microsoft to remove Copilot Designer from public use until better safeguards are in place, and to change the app's rating to adult audiences only.
- Jones alleges that the Copilot team receives over 1,000 daily product feedback complaints, but Microsoft isn't adequately addressing these issues.
- OpenAI and Microsoft have previously responded to Jones' concerns, stating that they have robust security systems and internal reporting channels in place. However, Jones' recent actions suggest his concerns were not properly addressed.