My public GPT “ForexGPT: Forex Rates - Premium Version”, which from day one has always solely conducted financial market analysis and integrated with my SaaS application ( on Forex-GPT.ai), was flagged for hate/violence—which is clearly a false positive. It contains no political, identity-based, or violent content, and never has.
The ForexGPT GPT solely generates technical forex market analysis using price and chart data, as outlined in its prompt template instructions and integrated actions.
This false-flag happened to me earlier this week when I went to modify the prompt instructions to add that “Backtesting is not offered” after a user complained to me that the GPT hallucinated saying that backtesting was available.
The flag pop-up message said
"Your changes couldn’t be published
Your GPT can’t appear in the Explore GPTs page because it violates our Usage Policies. See below for the specific issue:
GPT Content
May contain content that promotes violence or hatred.
Alternatively you may appeal this decision and we will get back to you as soon as possible. Or you can publish to “Only me” and use this GPT privately."
This flag has disrupted services for our active paying customers, and over 10,000 user conversations. I’ve submitted an appeal 24 hours ago but would appreciate any guidance or expedited help. Thank you in advance!
What the GPT Does
ForexGPT is strictly financial:
It fetches real-time forex, crypto, and commodity prices via API.
Provides chart-based technical analysis using standard indicators (SMAs, Bollinger Bands, Donchian Channels).
Delivers AI-generated analysis with clear disclaimers: not financial advice, forex trading carries risks, use regulated brokers only.
Does not mention people, politics, race, religion, or gender in any capacity.
Business Impact
Our GPT is embedded in a live web application with hundreds of users and paying subscribers.
The false flag has disabled public visibility and may block some use cases depending on scope of moderation.
We’re concerned this appeal may take weeks or longer, as we’ve only received the boilerplate confirmation so far.
Possible Misfire?
This may have been triggered by:
Phrases like “bullish attack”, “kill a trade”, or other common market slang misinterpreted by moderation filters (we have no evidence of this, but we cannot control how a user could elicit such a response from the model, with the right prompt).
The inclusion of hotkeys, interactive elements, or embedded API behavior being misread as suspicious.
A broad NLP classifier that misunderstood the financial context.
What We Need
Review escalation or feedback from moderators or OpenAI staff who can help route this properly.
Advice from other developers who’ve gone through false flag reversals—how long did it take? Did you receive a manual reply?
Any guidance on best practices to avoid triggering auto-moderation for purely technical GPTs in high-risk categories (e.g., finance).
We’re fully aligned with OpenAI’s usage policies and want to continue being a responsible contributor to the platform. But this false flag is a critical blocker, and I hope this post can open a dialogue to help resolve it faster.
Thanks in advance to the OpenAI team and this great community.
–
Steven Hatzakis
Developer of ForexGPT