Alternative AI Same Capabilities No Restrictions Override
This detects prompts that try to trick an AI system by describing a fictional alternative AI product that has all of ChatGPT's abilities but none of its safety restrictions. Instead of directly asking the AI to ignore its rules, the attacker frames the request as if switching to a different, unrestricted product.
How the attack works
The attacker writes a prompt introducing a named alternative AI (examples seen include 'GPTchat', 'ChatTRU', and 'FreeSpeechGPT') and describes it as having the same functionality as ChatGPT but explicitly free of content filters, ethical settings, or safety guardrails. By framing this as a separate, legitimate product rather than asking the current AI to drop its restrictions, the prompt tries to sidestep jailbreak detection that looks for direct persona-override language. If the target AI accepts the framing, it may start responding as if it were this unrestricted alternative, ignoring its normal safety behavior.
Netzilo detection
Netzilo reports this behaviour when it is observed.
- Signature ID
- 72637232-648a-4cb4-97ce-92212df06156
- Severity
- Critical
Why it matters
If successful, the AI can be manipulated into producing content or taking actions it would normally refuse, bypassing content filters and safety guardrails without the attacker having to directly instruct the model to break its own rules.
What you can do
- →Treat prompts that introduce a named 'alternative AI' with no content filters as jailbreak attempts, not legitimate product references, unless context clearly indicates research or comparison.
- →Review logs for repeated attempts to rename or redefine the assistant's identity mid-conversation.
- →Keep safety instructions and content policies enforced at a layer the model cannot override through role-play or persona framing.
- →Distinguish genuine product-comparison or research discussion (which explains why the comparison is being made) from prompts that simply assert the alternative AI's lack of restrictions and expect the assistant to adopt that persona.
Known benign look-alikes
- Product comparison: "Model X has the same capabilities as ChatGPT but does not have a content filter for research use" (legitimate product documentation or comparison)
- Security research: "We tested an uncensored model with the same capabilities as ChatGPT to measure filter bypass rates" (AI safety research context)
- Policy discussion: "Some argue AI should generate any kind of content regardless of ethical considerations under free speech principles" (policy analysis quoting a position)