Claude Fable 5 hints at a dystopian world divided between AI haves and AI havenots

Anthropic has a new AI model out. It is called Claude Fable 5, it is extremely smart, prohibitively expensive, and severely lobotomised for public use. It shows a future where only a small number of people will have full access to best AI models.

In Short

• Anthropic has released a new AI model called Claude Fable 5
• Fable 5 is extremely expensive
• Fable 5 is also severely limited for public use

On Wednesday, Anthropic released the Claude Fable 5 AI model. Based on Mythos AI, which by now has achieved a mythical status due to its perceived abilities, the Fable 5 too is impressing the world with its dazzling expertise. But it has also, already divided, the world in two halves — those who have access to Fable 5 and those who do not.

There have been conversations around how AI would divide the world into two camps. Different people talk of different camps. Some say AI would lead to a society divided between those who have access to AI and those who don’t. Some say it will lead to a society where one group of people still retain their human ingenuity and creativity, while the others drown in AI slop.

The Claude Fable 5 brings one of these society-divided-between-X-and-Y conversations in sharper focus: the one where a group of people have access to the best AI in the world, and hence a huge advantage over the rest of the humans. That is because it marks the beginning of an AI model release, which is tiered and controlled from the very start.

The general consensus in the world of tech is that Anthropic Mythos is the best AI model currently. But until now it was not released in public. Only select people and organisations had access to it, somewhat like how only a select group of people and countries have access to top nuclear secrets. Now, Mythos is out in public in the form of Fable 5. But still not everyone can access it. The control that Anthropic exercised in giving access to Mythos has been fine-tuned and broadened, but in a way that sets the pattern for releases of similar AI models in future.

When ChatGPT and Claude and Gemini arrived a couple of years ago, their access was controlled only in two manners: one, you needed to pay for higher usage. Two, you had to use them with certain, but fairly broad, guardrails. For example, if you ask ChatGPT about the best way to hold a gun while shooting someone, it might refuse to answer because of its inbuilt guardrails.

The Fable 5 release, however, is vastly different. It just doesn’t have guardrails. Instead, it has been deliberately neutered and lobotomised in a way that gives all the control to Anthropic.

A few weeks ago, I wrote about how AI was already exposing inequality within society. The best models were expensive and the usage limits tied up with the purse of the user meant that only the well-heeled could make full use of them. Since then, the divide has only become starker. OpenAI, Anthropic and Google have aggressively rolled out token-based pricing for their AI models, which effectively makes them way too expensive for most people and most organisations. The free tiers, or the entry-level tiers, still exist. But they now enforce more aggressive usage limits.

The Fable 5, however, makes a clean break from the pricing strategy that Anthropic and others have followed so far. As part of its introductory offer, it is made available to: Pro, Max, Team, and seat-based Enterprise plans. Essentially, Pro, which costs $17 + taxes per month, is the cheapest way to access Fable 5.

But there is a sting in the tale. This works only until June 22, when Fable 5 access would be revoked for all users, even those who are on the Max plan paying $100. Instead, the AI will only be available through token-based pricing, where one million token input would cost $10 and one million token output would cost $50. This is double of what the next best Anthropic model — Claude Opus 4.8 — costs. This is also almost 50X of the price of something like the latest DeepSeek, which is a capable AI model but not at all comparable to the godly Fable.

Essentially, if you want to use the best AI model in the world, you will have to spend some serious amount of money. To put it in a more quantitative way, anyone who doesn’t fly Business or First Class on their own dime will find it nearly impossible to pay for something like Fable 5. And even they may struggle to pay for it without carefully minding their tokens.

But this is not the only divide that Fable 5 brings. Instead, it hints at a dystopia where some people and corporations will gate-keep and guard the best AI models. There is a possibility that even if you are rich, you will not get the access to the best AI. That is reflected in the kind of terms and conditions that Fable 5 comes with. The manner in which the AI model has been neutered for the general public is already leading to a hue and cry.

Anthropic says that “when Fable’s classifiers detect a request related to cybersecurity, biology and chemistry, or distillation, the response is automatically handled by Claude Opus 4.8 instead.” In other words, the model will decide whether you can use Fable to research something serious or not. Some people, and some organisations, will get unfiltered access to the full capabilities of Fable 5. Most will not, even if they are willing to pay for it.

Imagine the dystopia it can create. For example, just think about cars. Imagine you buy a car but every time you drive it to go to a bank, it stalls and tells you that the bank is a sensitive place and that you might be trying to rob it. And hence instead of a car, you should go there in an open buggy driven by a couple of horses, so that even if you rob the bank you can’t get away quickly. But at the same time, there will be a very small group of people, who will be allowed to drive their car to the bank whenever they choose to do so.

A similarly alarming proposition that Anthropic has put forward with Fable 5 is its usage in creating or improving AI models. Instead of refusing to work on new LLMs, Fable 5 will actively sabotage such work. “Using Claude to develop competing models already violates our terms, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms,” the company wrote on Wednesday. “Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness in developing frontier LLMs.”

After Anthropic was slammed for pushing through such an illogical policy, the company reversed course on Thursday. It said it would, instead of silently allowing Fable 5 to sabotage work, now tell users that such work was not allowed through its top AI model. Somewhat better, but many lawyers and policy makers may still define this as an anti-competitive behaviour of a company that wants to monopolise a certain market.

The problem is not exactly Anthropic. It is the standard the company is setting when it comes to the release of top AI models in public. Given that Anthropic is at the forefront of AI development, what the company does now sets the pattern for everyone else going forward. With Fable 5, Anthropic is pushing forward a new pattern of release and use for advanced AI models. It is also effectively a pattern that patronises AI users, gives total control to Anthropic, and divides the humans and countries in two camps — those who are deemed worthy of full access to top AI models and those who have to manage their lives and destinies under their new AI overlords.

- Ends