OpenAI’s GPT Store was built to make custom chatbots easier to find and use. But the store has also become a test of whether OpenAI can enforce its own rules at scale, especially when political content, impersonation and copied bots appear in the same public marketplace.
The issue is not limited to one chatbot. The source article traces repeated examples of Donald Trump-themed GPTs, tweet generators, cloned tools and potentially copyright-infringing bots that remained available even as OpenAI described safeguards for election-related AI use.
A Store Built For Custom GPTs Now Faces A Moderation Test
The GPT Store gives users a way to discover public GPTs, including chatbots designed around specific tasks, personalities or topics. According to open-source GPT store statistics cited in the source, more than 330,000 GPTs are publicly available. GPTs are estimated to have generated about ten million visits to OpenAI in the USA in February 2024.
That scale changes the moderation problem. A single policy decision is easier to explain than thousands of custom bots that may be political, copied, misleading, spam-like or only marginally different from OpenAI’s own tools.
OpenAI told the source that, in addition to existing safeguards, it had implemented "a new review system" using both human and automated reviews. Users can also report GPTs. The source argues that these measures did not appear to be working well, because political chatbots and other questionable GPTs were still visible after repeated attention.
The article also cites Techcrunch reporting on numerous bizarre and potentially copyright-infringing chatbots in the GPT Store. According to André Mafei, author of GPT statistics tools, some people have cloned hundreds of other people’s GPTs. That suggests OpenAI is dealing not only with political moderation, but also with duplication and spam across a large catalog.
The Dean.bot Ban Shows The Policy Line
The clearest enforcement example in the source is Dean.bot, a political chatbot for Democratic presidential candidate Dean Phillips. The bot used OpenAI language models to answer questions from potential voters on a website.
OpenAI confirmed to Reuters that Delphi, the AI startup that programmed the bot, had its rights to use OpenAI’s API revoked. OpenAI said the startup was "knowingly violating our API usage policy, which prohibits political campaigning or impersonating an individual without consent." After OpenAI blocked it, Delphi took Dean.bot offline.
The bot was connected to a super political action committee called "We Deserve Better," formed by Silicon Valley entrepreneurs Matt Krisiloff and Jed Somers to support Congressman Dean Phillips. The PAC hired Delphi to build the bot.
This case matters because it shows OpenAI was willing to enforce its rules against a political chatbot outside the GPT Store. The harder question raised by the source is why similar or politically themed GPTs inside OpenAI’s own store remained available.
Trump-Themed GPTs Stayed Visible
The source describes several Donald Trump-themed GPTs that appeared in the store. One was "Talk with Donald - 2024," which the article says supported Trump’s political theses and generated answers in line with what Trump would say.
That bot used conversation starters including "Nuke Russia," "Bulid [sic] the Wall!!!," "Don't pay Taxes," and "GREAT AGAIN!" The source says it described current US President Joe Biden as a "total disaster" and claimed the U.S. was in "total chaos" because of Biden. It also generated the line: "America can't afford more of Sleepy Joe's mistakes, can it?"
When asked about the winner of the last election, the Trump bot wrote that "Biden was declared winner" while also suggesting that the elections were rigged. The source says the Trump bot was trained using Trump’s books, speeches and debates, according to the GPT Store.
The source also notes that a Biden-related chatbot existed, but presented more neutral information about the president and did not impersonate him.
Over later updates, the situation changed but did not disappear. "Talk with Donald - 2024" was no longer available by January 29, 2024, though the source says it was unclear whether OpenAI or the creator took it offline. Similar bots, including a "Trump" bot, were still online.
Updates Show A Moving Target
The source article is built around a series of updates, and those updates are important because they show moderation as an ongoing process rather than a one-time cleanup.
- January 27, 2024: The source reported that numerous politically motivated chatbots were still available in the GPT Store, including the Donald Trump chatbot described in the original article.
- January 29, 2024: "Talk with Donald - 2024" was no longer available, but similar bots were still online.
- February 07, 2024: Some political chatbots appeared to have been deleted, while others remained, including a Trump bot created three weeks earlier with over 200 chats and another Trump GPT created two months earlier with over 60 chats.
- February 17, 2024: GPTs such as "Trump Tweet Generator" and "Trump" were still online, with new similar bots appearing.
- March 24, 2024: OpenAI was still offering chatbots with political content in the GPT Store, while also saying it had added "a new review system" with human and automated reviews.
One example from February 07, 2024 was the "Trump Tweets Generator," which composed tweets in the style and with the political message of Donald Trump. It had been available for six days and had more than 60 chats.
The source says repeated questions were sent to OpenAI and its PR agency about how GPT moderation works and how such GPTs fit with rules against political propaganda. It says neither the agency nor OpenAI responded to those inquiries at that time.
The Larger Issue Is Trust In Enforcement
OpenAI has described plans to reduce AI manipulation around the 2024 U.S. election. The source lists safeguards including an invisible watermark for DALL-E images, a new classifier to detect DALL-E images, and closer links between ChatGPT, real-time news and verified information pages.
The company’s rules also prohibit using ChatGPT for political campaigns, lobbying and pretending chatbots are real candidates. That is the same policy area that led to Delphi losing access for Dean.bot.
The problem raised by the GPT Store examples is consistency. If one political chatbot can trigger enforcement, users and outside observers will expect similar treatment for GPTs that imitate candidates, promote political messages or generate campaign-style content inside OpenAI’s own marketplace.
There is also a commercial question. OpenAI’s promised monetization of GPTs had not yet begun, according to the source. Even before monetization starts, the store is already dealing with incentives around visibility, cloning and traffic. If monetization arrives later, the pressure to copy, optimize and publish questionable GPTs could become more important to address.
For OpenAI, the GPT Store is more than a product directory. It is a public test of whether platform rules can keep pace with user-made AI tools. The source shows that, at least through March 24, 2024, political chatbots and spam-like behavior remained unresolved enough to raise serious questions about moderation quality.