OpenAI revealed on Friday that an alleged technical issue was the reason behind ChatGPT mistakenly flagging safety warnings on links to WinRed, the primary online fundraising platform for Republican candidates. Notably, ActBlue, the main Democratic fundraising platform, did not trigger a similar warning.
The problem came to light when digital marketer Mike Morrison noticed the discrepancy while requesting links from WinRed and ActBlue from ChatGPT. Morrison shared his findings on social media, pointing out that WinRed links were marked as potentially unsafe while ActBlue links were deemed safe.
When ChatGPT provided links to GOP-affiliated stores on WinRed, it included a warning advising users to verify the safety of the link and mentioning potential data sharing with third-party websites. This warning was not applied to links from the Democrat fundraising platform, as per Morrison’s observations.
WinRed CEO Ryan Lyk criticized the evident bias, labeling it as “election interference.” OpenAI quickly responded to the situation, acknowledging the issue and stating that it was being addressed.
OpenAI’s spokesperson clarified that the issue was not politically motivated but rather a result of the model generating website links that were not in the search index. They assured that the problem was being resolved, attributing it to how URLs were discovered.
In conclusion, the incident involving ChatGPT flagging GOP websites as potentially unsafe due to a technical glitch was swiftly addressed by OpenAI, with measures being taken to resolve the issue.