Anthropic Previews Full Release of Super Model Claude Mythos Within Weeks! Hacker-Grade AI About to Be Unleashed

📄Full Article· Automatically extracted by trafilaturaGemini 翻譯1667 words

The ultimate double-edged sword of AI cybersecurity is about to be unsheathed! According to Reuters, Anthropic has officially confirmed that "Mythos," a super AI model with top-tier cyber-hacking capabilities that has drawn intense attention from global leaders, is expected to be released to customers within the next few weeks. Previously, the model was strictly controlled due to its ability to autonomously discover software vulnerabilities, and was only made available to giants such as Apple, Microsoft, and Amazon for defensive testing through a special project. Meanwhile, Anthropic also officially launched the upgraded Claude Opus 4.8 today (the 28th), featuring stronger autonomous capabilities and improved model "honesty." (Previously: Anthropic launches the latest model Claude Opus 4.8! Claude Code simultaneously adds "Dynamic Workflows," one person can match a development team of a hundred) (Background: Anthropic's revenue overtakes OpenAI: a historic reversal in the AI arms race) The global AI compute arms race is about to welcome a heavyweight disruptor. According to a Reuters report today (May 28, 2026), the well-known AI unicorn Anthropic announced that, although it has already rushed out the upgraded Claude Opus 4.8 today, the company is currently going all out to prepare the market-shaking super model Claude Mythos for official release to all customers "within the next few weeks." The Strongest Hacker AI Unleashed: Mythos and Project Glasswing As early as the beginning of April this year, Anthropic first publicly previewed the existence of the Mythos-class model. This is the most powerful Frontier model the lab has ever developed, achieving a massive leap in general capabilities. However, what makes Mythos most eye-catching (and at the same time most concerning) is that it has reached expert-level "cybersecurity and hacking capabilities." It is said that the model can autonomously find and exploit Zero-days in major operating systems, browsers, and software. This highly lethal double-edged-sword characteristic has triggered strong concerns from corporate executives and national leaders worldwide about the possibility of it being maliciously weaponized. To ensure safety, Mythos had previously been held in a strict "Restricted" state and was not opened to the general public. Currently, Anthropic only allows rigorously vetted tech giants — including Amazon (AWS), Microsoft, Apple, Google, and cybersecurity heavyweight CrowdStrike, among others — to use Mythos for system vulnerability testing and software hardening through a defensive project called Project Glasswing. As additional safety safeguards have been put in place, Anthropic has now finally nodded in approval, preparing to gradually open this performance behemoth to enterprises and customers that pass review. Bonus Feature: Opus 4.8, Marketed on "Honesty," Goes Live Today While the outside world eagerly awaits the full unlocking of Mythos, Anthropic has also delivered a sincere transitional weapon: Claude Opus 4.8 officially went live today. Opus 4.8 continues Anthropic's strategy of steady updates, maintaining exactly the same pricing as its predecessor 4.7 ($5 / $25 per million input/output tokens), and brings several key upgrades: - Cheaper Fast Mode: A new Fast Mode that is 2.5x faster has been added, at roughly one-third the price of the previous fast mode, making it extremely well-suited for high-throughput development work. - Performance and Agent Upgrades: Better performance on software engineering benchmarks such as SWE-Bench and Terminal-Bench, with support for "Dynamic workflows" that can launch hundreds of parallel sub-agents simultaneously. - Significantly Improved "Honesty": This is the biggest highlight of Opus 4.8. Officials note that early testers have reported that the model is now more inclined to flag uncertainty in its own work, greatly reducing the probability of confidently producing wrong conclusions without evidence (i.e., AI hallucinations), and its Alignment performance is now very close to that of the Mythos preview version. With Opus 4.8 in place and Mythos powerfully arriving in just a few weeks, Anthropic is clearly further expanding its dominance in the enterprise AI market through a dual-track strategy that runs "safety upgrades" and "ultimate performance" in parallel.

Data Status✓ Full text extractedRead Original (動區 BlockTempo)

🔍Historical Similar Events· Keyword + Asset Matching6 items

2026-05-28

Anthropic's Claude Mythos AI Model Nearing Release After Raising Cybersecurity Alarms

2026-05-14

Apple Mac M5 System Exploited With Anthropic's Claude Mythos AI, Researchers Claim

2026-05-03

GPT-5.5 security testing scores are nearly identical to Claude Mythos, with the latest evaluation debunking Anthropic's claim that it is too dangerous to be released.

2026-05-03

GPT-5.5 security testing scores are nearly identical to Claude Mythos, debunking Anthropic's claim that it is too dangerous to be released.

2026-05-01

OpenAI's GPT-5.5 Matches Claude Mythos in Cyberattack Capabilities: AI Security Institute