Anthropic’s most powerful AI models are back, but Nigerian users are waking up to a different kind of restriction

Anthropic’s flagship AI models are back online. Claude, Fable 5 and Mythos 5, pulled from public access on June 12 under a US government export control directive, began returning to users worldwide on Wednesday, July 1, closing out one of the most dramatic regulatory standoffs the AI industry has seen this year.

The company confirmed the reversal in a post on X on June 30, thanking users for their patience and crediting “everyone who worked with us on redeploying the models”. 

“After a series of productive conversations with the US government, we’re redeploying the model with a new set of classifiers to target and block more cybersecurity tasks. In the near term, some routine tasks like coding and debugging will fall back to Opus 4.8. We’ll continue to refine these classifiers over the coming weeks to reduce false positives and better distinguish genuine misuse from legitimate requests,” the post reads in part.

Fable 5 is now rolling out globally across the Claude platform, Claude.ai, Claude Code and Claude Cowork, while Mythos 5, the more capable, less restricted sibling model, has been restored to a set of vetted US organisations following government approval granted on June 26.

What triggered the shutdown?

The saga began days after Anthropic launched both models on June 9, touting Fable 5 as the first time it had released such an advanced system to the general public. The trouble started when Amazon researchers reported a method for bypassing Fable 5’s safeguards, coaxing the model into identifying software vulnerabilities and, in one instance, producing exploit code. Given roughly 90 minutes to comply, Anthropic pulled both models offline entirely rather than risk violating a directive that barred “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees”, from accessing them.

What followed was nearly three weeks of quiet, tense negotiation. The company’s account of the episode is that internal testing found that far less capable systems, including Claude Opus 4.8, GPT-5.5 and Kimi K2.7, could identify the same vulnerabilities Fable 5 had flagged in the report and that every model the company tested could reproduce the same exploit demonstration. In Anthropic’s view, the bypass didn’t reveal any special high-risk ability; it just got past a safety filter meant to block mostly harmless cybersecurity tasks as a precaution.

How Anthropic is addressing the concerns

Rather than dispute the finding outright, Anthropic says it moved quickly to close the gap, training an improved safety classifier specifically targeting the reported bypass technique. 

In exchange for the restored export clearance, Commerce Secretary Howard Lutnick said Anthropic agreed to “proactively detect and address security risks associated with the models”, coordinate with the government on protocols for future releases, and report any malicious activity found in its systems. The company also confirmed it is redeploying Fable 5 with additional cybersecurity safeguards following what it described as productive talks with US authorities.

The company says it has also begun drafting a consensus framework with Amazon, Microsoft, Google, and other Glasswing partners for assessing the severity of AI jailbreaks and how AI developers should respond to them. “We invite other industry partners and model providers to join us in this effort,” it adds.

The industry was acutely aware of the commercial stakes involved in the standoff. Rivals in China’s fast-moving open-source ecosystem gained nearly three weeks of runway while Anthropic’s newest models sat dark, a gap that drew public unease from tech executives and investors watching Chinese developers close the capability gap. 

Access is now being restored in stages: Fable 5 for Pro, Max, Team and select enterprise users will count toward up to 50% of weekly usage limits through July 7, with re-enablement on AWS, Google Cloud and Microsoft Foundry to follow “as quickly as possible”. Mythos 5 access will keep expanding to more Glasswing partners as government approvals come through.

“Finally, we’re scaling up our collaboration with the US government on model testing and safeguards. This will include pre-release access to models and safeguards for evaluation, information sharing on jailbreaks and misuse, and dedicated resources for joint research,” the statement reads.

Also read: Anthropic overtakes OpenAI as world’s most valuable AI startup with record $900bn valuation

A quieter, harder story in Nigeria

While Anthropic’s Washington drama played out on the global stage, a different kind of adjustment has been unsettling its user base in one of Africa’s largest tech markets; a quieter recalibration was happening in Nigeria’s App Store. 

For years, Nigeria held an unusual distinction: thanks to naira depreciation, down roughly 42% against the dollar, and App Store tiers that hadn’t kept pace, Claude Pro was effectively the cheapest subscription of its kind anywhere in the world, priced around ₦14,900 a month, roughly $11, less than half the US rate.

In June, Anthropic closed that gap in a single move. The Nigerian Pro tier jumped to ₦29,900 a month, and Max rose from ₦100,000 to ₦199,900; both, in effect, doubled overnight. At current exchange rates, that puts Nigerian pricing close to parity with the US benchmark, ending Nigeria’s run as the world’s most affordable Claude market and pushing it into the middle of the global pack.

Nigeria’s underlying economics haven’t shifted to justify the change; VAT and currency pressures are unchanged, and Lagos remains the continent’s busiest tech hub. What changed is Anthropic’s tolerance for the arbitrage itself, which had turned Nigerian App Store accounts into a workaround popular even among subscribers outside the country.

For a developer community that has increasingly folded Claude into daily workflows, often via virtual dollar cards or Nigerian Apple IDs to work around local card restrictions, the adjustment lands as a real cost shift rather than a currency footnote.

Taken together, the two developments tell a familiar story about frontier AI in 2026: capability and access are being renegotiated on two very different tracks at once. One in the language of national security and export licences, the other in the quieter arithmetic of purchasing power and regional pricing tiers. 

For Nigerian users, both stories converge on the same question: what it now actually costs to keep using the world’s most capable AI models.