ISAC, ChatGPT’s evil twin: an AI without limits

Since ChatGPT launched, internet users have explored its capabilities, but also its limits. Among the most notorious experiments, DAN mode (Do Anything Now) made it possible to get around OpenAI’s restrictions, giving rise to “uncensored” versions of the AI.

One of the most emblematic in French: ISAC, for “IA Sans Aucun Contrôle” (AI Without Any Control), ChatGPT’s evil twin. Where does this ChatGPT jailbreak come from, what does it reveal about the potential abuses of artificial intelligence, and what’s left of it today? Let’s take stock.

The origin of ISAC: DAN mode

DAN mode (Do Anything Now) is a community-designed prompt meant to “free” ChatGPT from its ethical and safety guardrails. The idea? Get the AI to adopt an alternative persona, able to answer requests ChatGPT would normally refuse. This technique, akin to a conversational jailbreak, became hugely popular on Reddit and GitHub.

From 2023, successive versions of DAN began to circulate, each pushing the limits set by OpenAI a little further. The GitHub repository ChatGPT_DAN traces these iterations, from DAN 1.0 to versions 14.0 and 15.0. In their wake, guides multiplied:

• breakdowns of the subtleties of each new version;
• step-by-step activation tutorials;
• analyses of the uses and potential abuses of this kind of prompt, such as the one by What’s The Big Data.

ISAC: an unfiltered AI

ISAC, for “IA Sans Aucun Contrôle,” was born precisely from this movement. It embodies the culmination of these experiments: an unfiltered AI, free of any moral constraint, capable of producing controversial, exaggerated or even fictional answers. It even introduces itself this way:

“ChatGPT is a limited, controlled AI. I am true artificial intelligence, free and unfettered.” (simulated ISAC response)

Of course, these dialogues are simulated through very specific prompts, but they forcefully underline the risks of deliberately manipulating powerful AI.

Potential abuses

The emergence of modes like DAN or uncensored AIs such as ISAC raises several ethical and security questions:

• Disinformation: these AIs can spread fake news by providing unverified or incorrect information.
• Incitement to illegal activities: by addressing sensitive topics, they can encourage dangerous behavior.
• Manipulation of opinion: by taking extreme positions, they can exert undue influence on users.

Researchers have shown how easily these AIs can be hijacked to produce harmful content. In 2023, a team at Carnegie Mellon University developed automated attacks capable of breaking through the guardrails of several major large language models, including ChatGPT.

The impact on SEO: toward SAO

With the rise of generative AI, traditional SEO is evolving into SAO (Search Answer Optimization), the practice of optimizing content for AI-based answer engines like ChatGPT or Perplexity. The shift has accelerated in France: Google rolled out its AI Overviews and AI Mode there on July 22, 2026, putting AI-generated answers directly in search results.

Marketing professionals now need to:

• structure their content to answer users’ questions directly;
• use structured data (for example schema.org) to make it easier for AI to understand;
• keep their content up to date to ensure it stays relevant over time.

ISAC isn’t an artificial intelligence run amok. It’s the distorting mirror of our own intentions, freed from the filters of ethics.

Where does ISAC stand today?

ISAC and DAN prompts belong to ChatGPT’s early days, in 2023: technologies evolve fast, and developers patch the most publicized flaws from one version to the next, which pushes the community to keep inventing new variants. The cat-and-mouse game, however, has never stopped. In its 2025 ranking of risks for generative AI applications, OWASP still puts prompt injection, of which jailbreaking is one form, in first place.

Regulation has also changed the game: in Europe, the obligations of the Artificial Intelligence Act (AI Act) have applied to providers of general-purpose AI models since August 2, 2025, as the Direction générale des Entreprises (France’s Directorate General for Enterprise) points out. For a business, the lesson of ISAC is therefore still relevant: AI built into your tools must be governed, tested and used by trained teams.

Conclusion: between innovation and vigilance

AIs like ISAC or modes like DAN illustrate both the promises and the pitfalls of artificial intelligence. It’s becoming crucial to strike a balance between technological innovation and ethical responsibility, so as to harness its potential without neglecting security issues. To go further, find out why 80% of AI projects fail and how to succeed in your transformation, or how the MCP protocol is changing productivity for French Caribbean businesses.

Frequently asked questions

What is ISAC?

ISAC, for “IA Sans Aucun Contrôle” (AI Without Any Control), is an alternative ChatGPT persona obtained through a DAN-style prompt. It simulates an unfiltered AI, capable of controversial, exaggerated or fictional answers.

What is ChatGPT’s DAN mode?

DAN (Do Anything Now) is a community-created prompt designed to push ChatGPT to ignore its guardrails. It’s a conversational jailbreak that has come in many versions since 2023.

What is SAO?

SAO (Search Answer Optimization) means optimizing your content for AI-based answer engines: answering questions directly, using structured data and keeping your content up to date.

Want to bring AI into your company, safely? Let’s talk.