Anthropic previews that the super model Claude Mythos will be fully opened to the public within weeks! Hacker-grade AI is about to have its restrictions lifted

The ultimate double-edged sword of AI cybersecurity is about to be unleashed! According to Reuters, Anthropic officials have confirmed that the highly anticipated super AI model “Mythos,” which has top-tier hacking capabilities and has drawn intense global attention from executives, is expected to be made available to customers within the next few weeks. Previously, the model was tightly restricted because of its ability to autonomously identify software vulnerabilities, and it was only made available to defense-focused testing projects run by major companies such as Apple, Microsoft, and Amazon. Meanwhile, today (28), Anthropic has also officially launched the upgraded Claude Opus 4.8, featuring stronger autonomous capabilities and the model’s “honesty.”
(Background recap: Anthropic releases the latest model Claude Opus 4.8! Claude Code simultaneously adds “Dynamic Workflows,” letting one person replace a hundred developers)
(Additional background: Anthropic’s revenue surpasses OpenAI—a historic turnaround in the AI arms race)

The global AI computing power arms race is about to welcome a heavyweight disruptor.

According to a Reuters report dated May 28, 2026, well-known AI unicorn Anthropic announced that, although it rolled out the upgraded Claude Opus 4.8 ahead of schedule today, the company is currently pushing full speed, preparing to officially release the market-shaking super model Claude Mythos to all customers “within the next few weeks.”

Unleashing the strongest hacker AI: Mythos and Project Glasswing

As early as the beginning of April this year, Anthropic first publicly tipped the existence of a Mythos-class model. This is the most powerful frontier model (Frontier model) the lab has developed to date, representing a huge leap in general-purpose capabilities.

However, what makes Mythos most striking (and also most concerning) is that it reaches expert-level “cybersecurity and hacking capabilities.” It is reported that the model can autonomously find and exploit zero-day vulnerabilities (Zero-days) in major operating systems, browsers, and software. This extremely lethal double-edged-sword trait has triggered intense concerns among corporate executives around the world and leaders in various countries about its potential to be weaponized for malicious use.

To ensure safety, Mythos has previously remained under strict “Restricted” internal controls and has not been opened to the general public. At present, Anthropic only allows selected tech giants—including Amazon (AWS), Microsoft, Apple, Google, and the cybersecurity firm CrowdStrike—to use Mythos for system vulnerability testing and software hardening through a defensive project called Project Glasswing. With additional security protection mechanisms gradually in place, Anthropic now finally gives the green light and is preparing to gradually open this performance powerhouse to vetted enterprises and clients.

Also featured: Opus 4.8 focused on “honesty” goes live today

While the outside world has been waiting for Mythos’s full release to be rolled out, Anthropic has also brought out a sincere transitional weapon: Claude Opus 4.8 is officially launching today.

Opus 4.8 continues Anthropic’s steady update strategy, keeping the exact same pricing as its predecessor 4.7 (for every 1 million input/output tokens: 5 USD / 25 USD), and brings several key upgrades:

• Cheaper Fast Mode: A fast mode has been added that is 2.5 times faster, and its price is about one-third of the previous fast mode—ideal for high-throughput development work.
• Performance and agent upgrades: Performs better in software engineering tests such as SWE-Bench and Terminal-Bench, and supports “Dynamic workflows” that can simultaneously launch hundreds of parallel sub-agents.
• Significantly improved “Honesty”: This is Opus 4.8’s biggest highlight. The official notes that feedback from early testers indicates the model is now more likely to flag the uncertainty in its own work, greatly reducing the chance of confidently delivering wrong conclusions without evidence (i.e., AI hallucinations). Its alignment performance is now very close to the level of the Mythos preview version.

With Opus 4.8 in place and Mythos coming in strong in the final few weeks of the countdown, Anthropic is clearly pursuing a dual-track strategy of “security upgrades” and “extreme performance,” further expanding its dominance in the enterprise AI market.