The US government pulls the plug on Anthropic’s Fable 5!

In partnership with

Watch every match the way it was meant to sound.

This summer, 48 nations play. For a lot of fans in the US, the match doesn't feel right in the wrong language. The commentary, the energy, the way goals sound when your language is calling them.

Norton Neo is a free browser with a free built-in VPN. Stream privately in the language you want. No subscription, no sign-up, no credit card. Built-in VPN, anti-fingerprinting, and ad blocking, all backed by Norton security.

Free to download. Free to use.

Fast. Safe. Intelligent. That's Neo..

Download Norton Neo

Last week, we celebrated the launch of Anthropic’s most powerful public model to date. Even better, the internet was buzzing with people using Fable 5 to pull off some pretty impressive ops.

And this week? The US government pulled the plug. 

Forward thinkers, welcome to another issue of The Neural Frontier. 

Let’s unpack the Fable 5 dilemma, as well as headlines from Apple and Microsoft. 

In a rush? Here's your quick byte: 

🙅 The US government pulls the plug on Anthropic’s Fable 5!

💰 Apple might raise product prices because of AI.

🚀 Copilot Cowork is now generally available!

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week.

🙅 The US government pulls the plug on Anthropic’s Fable 5!

In one of the most dramatic moments in AI policy to date, Anthropic says it has been ordered by the US government to suspend access to its flagship models, Fable 5 and Mythos 5, for all users worldwide.

The directive reportedly arrived with almost no warning and immediately forced Anthropic to disable access for customers, developers, enterprises, and even foreign-national employees working inside the company.

⚖️ What happened?

According to Anthropic, the government invoked national security authorities after becoming concerned about a potential jailbreak technique that could bypass some of Fable 5's safety safeguards.

The company claims it was only shown evidence of a narrow technique that could identify a small number of previously known software vulnerabilities. Anthropic argues that those vulnerabilities are:

• Already known

• Relatively minor

• Discoverable by other publicly available models

• Not unique to Fable 5 or Mythos 5

The company says it has not been shown evidence of a universal jailbreak or any example of the model causing meaningful real-world harm.

🛡️ Anthropic's defense

A large portion of the statement is essentially Anthropic defending its safety strategy.

The company argues that perfect jailbreak resistance is currently impossible for any frontier AI system. Instead, its goal has been to make attacks difficult, expensive, narrow in scope, and detectable when they occur.

Anthropic says Fable underwent thousands of hours of red-teaming involving government agencies, third-party evaluators, internal teams, and the UK's AI Safety Institute before release.

The company also points to controversial safeguards it already implemented, including mandatory 30-day retention of customer activity, specifically so it could investigate and mitigate potential abuse.

🏛️ The policy fight underneath the story

What's really interesting is that Anthropic is not arguing against government oversight. In fact, the company explicitly says governments should have the authority to stop unsafe AI deployments.

What it is arguing against is the process. Anthropic says any intervention should be:

• Transparent

• Technically grounded

• Fairly applied across the industry

• Based on clear evidence

The company believes this directive fails those tests. That's a notable position because it puts Anthropic in the unusual position of simultaneously supporting stronger AI regulation while criticizing a specific regulatory action.

🌍 Why this matters beyond Anthropic

If the government's standard is that any model with a potentially exploitable jailbreak can be removed from the market, the implications extend far beyond Anthropic.

Every frontier AI company currently deals with jailbreaks. OpenAI, Google, Meta, xAI, DeepSeek, and others all continuously patch exploits, prompt attacks, and safety bypasses.

Anthropic's core argument is that if a narrow jailbreak is enough to trigger a recall, then effectively no frontier model would remain deployable.

That is why this may end up becoming one of the most important AI governance disputes we've seen so far.

💰 Apple might raise product prices because of AI.

For years, AI has been framed as a software revolution. Now it's creating a hardware crisis.

Apple is warning that the global shortage of memory and storage chips, "RAMageddon," if you will, may force price increases across its product lineup, potentially making future iPhones, Macs, and iPads significantly more expensive.

⚠️ The hidden cost of the AI boom

Every major AI model depends on enormous quantities of memory. Training frontier models requires it, running them requires it, and serving billions of AI requests every day requires even more of it.

As AI companies race to build larger models and larger data centers, they are consuming massive amounts of DRAM and NAND chips, the same components that power consumer electronics.

The result is a global supply squeeze that has pushed memory costs dramatically higher. According to Tim Cook, chip costs have increased roughly fourfold compared to last year.

Apple has absorbed those increases so far, but Cook now says that's becoming unsustainable.

📱 Why the iPhone is in the crosshairs

Apple hasn't announced specific price increases yet. But industry analysts believe the iPhone is the most likely place they'll show up. The timing is obvious.

Apple's next iPhone launch is expected in September, giving the company a natural opportunity to reset pricing while introducing new hardware.

Research firm TechInsights estimates Apple would need to add roughly $270 to the price of a future iPhone Pro to fully preserve its current margins. Whether Apple passes through the entire increase remains unlikely.

But the direction seems increasingly clear. Consumers are likely to absorb at least part of the AI infrastructure bill.

🧠 AI is creating a second problem for Apple

The irony is that Apple isn't just being hurt by AI infrastructure costs. It's also under pressure because of AI itself.

Unlike Microsoft, Google, Anthropic, OpenAI, or Meta, Apple has struggled to establish a convincing AI position. The company spent much of the past two years trying to catch up after missing early generative AI shifts.

That effort has included:

• A major Siri overhaul

• More on-device AI processing

• New AI features across iOS and macOS

• A $250 million settlement related to AI features it failed to deliver

The challenge is that many of these improvements require even more memory. In other words, the solution to Apple's AI problem may worsen its hardware cost problem.

🚀 Copilot Cowork is now generally available!

Three months ago, Microsoft introduced Copilot Cowork as an experimental glimpse into the future of work.

Now it's generally available worldwide, and Microsoft says more than half of the Fortune 500 have already used it during the preview period.

🧠 What makes Cowork different?

Most AI tools help you create something. Cowork is designed to finish something.

Instead of generating a draft, answering a question, or recommending next steps, Cowork takes ownership of an entire workflow and runs it end-to-end. Microsoft shared examples from early customers that highlight the shift:

• An engineering team automated spreadsheet updates and dependency mapping that previously required manual review.

• A product team compared nearly 4,000 files across two software versions in a task that would have taken weeks.

• A sales leader analyzed a stalled pipeline and received prioritized opportunities with recommended follow-ups in a single morning.

The common theme is delegation rather than assistance.

⚙️ Microsoft's strategy: leverage the enterprise advantage

While OpenAI, Anthropic, and Perplexity are building increasingly powerful agents, Microsoft's advantage is context.

Cowork runs directly inside the Microsoft 365 ecosystem, drawing from what Microsoft calls Work IQ, its system for understanding organizational data across Outlook, Teams, Excel, PowerPoint, SharePoint, and other enterprise systems.

That means Cowork isn't starting from scratch every time. It already understands your meetings, documents, emails, business processes, and organizational permissions

💰 The business model is becoming clearer

One of the most interesting aspects of the launch is Microsoft's pricing approach.

Rather than bundling unlimited usage into Microsoft 365 Copilot, Cowork introduces usage-based billing.

Organizations need a Microsoft 365 Copilot license first. After that, they pay separately for Cowork tasks based on model usage, context retrieval, tool calls, and runtime.

Microsoft is effectively treating agentic work like cloud infrastructure. The more complex the task, the more compute it consumes, and the more it costs.

⚡ The Neural Frontier’s weekly spotlight: 3 AI tools making the rounds this week.

1. 🧭 Waybound is an AI use-case audit tool that evaluates your workflows and tells you whether they're worth automating with AI.

2. 🦍 IdeaApe is an AI market research tool that mines real Reddit conversations to uncover authentic audience motivations, barriers, and trends, straight from people's own words.

3. 🎯 Caddie is an AI sales copilot that joins your calls live, instantly answering questions from your knowledge base and catching you the moment you stall, so deals stop leaking and start closing.

Wrapping up…

Topsy-turvy doesn’t even begin to describe the week we’ve had. From product releases to government bans, we always get the sense that there’s something up in the air. 

Whatever it is, you best believe that we’re always on it. And as usual, the latest will be right in your inbox, same time next week 👋!

PS: If you’re the “friend” this mail was forwarded to, and you enjoyed it, hit the Subscribe button to see more content like this every week 🙂.