Outside of the United Kingdom, entities in Europe have not been granted early access to the powerful new Mythos AI. That now appears to be changing, at least in a limited fashion, as a spokesperson for the European Commission has confirmed that Anthropic has agreed to let leading European Union cybersecurity organization ENISA have the same early access for testing purposes that select US and UK organizations already enjoy.
The move comes after weeks of concerns being aired about Europe being left behind in preparedness for the advanced capabilities the Mythos AI and other comparable competitors will have in identifying vulnerabilities, including a meeting in which the bloc’s banks were encouraged to drastically increase cybersecurity spend to make ready. On June 2, Anthropic confirmed that it will be adding 150 new organizations across some 15 countries to the “Project Glasswing” access to Mythos Preview and that ENISA will be among these.
Mythos AI to be available to an assortment of new countries
Launched in early April, and at first limited to just 40 participants, Project Glasswing is giving government agencies and prominent private organizations (those among the most high-value targets for hackers) the ability to test their cybersecurity against a preview of the forthcoming Mythos AI. The results thus far have shown that organizations can expect frontier models of this caliber to potentially uncover tens to hundreds of vulnerabilities in a very short time, including some that may have been buried for decades without prior detection. There is also the possibility of it autonomously assisting with the deployment of exploit chains, though there is less compelling evidence of that being a serious threat at present.
Uzair Gadit, CEO of Secure.com, highlights why this move was necessary for the region: “Giving a regulator like ENISA hands-on access to a frontier model is a smart move. Defenders learn fastest when they can test these systems directly, not read about them secondhand. This is a well considered move, aligning with Anthropic’s filing to go public.”
“Europe putting its own experts that close to the technology is how you build informed policy instead of guessing at it,” Gadit added. “The threat landscape didn’t evolve — it accelerated. What used to require a skilled hacker and days of preparation now takes an AI tool and 20 minutes.”
ENISA access was first reported to the press by Tech Sovereignty Thomas Regnier of the European Commission, who confirmed that the cabinet had “several productive meetings” with Anthropic in recent weeks. The current participants have been given access to a pool of $100 million in access credits for testing, but this number will almost certainly be increased given the amount of new organizations coming on board.
Anthropic has not set a firm release date for the new Mythos AI, and has sent some confusing mixed signals as of late. An announcement for the recently released Opus 4.8 included a small segment that indicated Mythos was coming to “all customers” in the “coming weeks,” just after another fairly recent statement that it would likely not be broadly available for at least 6 to 12 months but that comparably powerful competitors would likely launch in the meantime. That announcement has caused prediction markets to strongly favor Mythos being launched in July, but industry analysts tend to believe it will not be available for enterprise API access until later in 2026 and for more general consumers until 2027.
Joshua Marpet, Senior product security consultant at Finite State, speculates about what discoveries might take place during whatever waiting period still remains: “Mythos, while reportedly equaled in capability by ChatGPT 5.5, among other frontier models, is still an incredibly powerful AI framework. The usage of Mythos by ENISA is fascinating. Will they use it to find vulnerabilities in EU RED and EU CRA Certified products? Or products coming up for certification? Are they going to try to use it to determine what exploits should be rated at what level? I have to assume that there are multiple questions that Mythos can and will answer for ENISA. Importantly, will this change the initial certification or certification maintenance process? That’s a question to be answered in the fullness of time.”
New Glasswing participants must meet Anthropic security requirements
Thus far, Anthropic says that Mythos AI has discovered over 10,000 real-world vulnerabilities in tester systems. In addition to the US government the project includes the big names of the tech world, including AI development rivals such as Microsoft, as well as major cybersecurity firms and members of the financial industry. Anthropic has not yet released a full list of the 150 new organizations that are soon to be included in Project Glasswing, but some have independently confirmed that they will be added; these include Samsung, authentication giant Okta, NATO, South Korea’s Korea Internet & Security Agency (KISA), and South Korean phone and internet service leader SK Telecom.
Mythos AI is considered to be on the leading edge of frontier AI models that both pose novel cybersecurity threats and will likely be available near-term to threat actors. The other models in this pot are OpenAI’s GPT-5.5-Cyber, and Google’s “Big Sleep.” OpenAI also recently promised EU entities access to the advance version of GPT-5.5-Cyber for similar testing. Google is not currently running a program similar to Project Glasswing, but has said that Big Sleep has autonomously detected and reproduced numerous security vulnerabilities in open source projects and had identified at least one that had previously only been known to threat actors.
John Carberry, Solution Sleuth at Xcape, thinks that the security community already has to regard traditional manual patching as “obsolete” despite these models not yet being available: “Project Glasswing’s integration of the European Union Agency for Cybersecurity, or ENISA, represents a critical geopolitical rebalancing, allowing international defenders to scan critical infrastructure before adversarial actors weaponize those same flaws. For security leaders, this transition means traditional, human-centric patching timelines are officially obsolete, shifting the enterprise bottleneck from vulnerability discovery to human remediation capacity. Organizations must proactively integrate automated code review, implement machine-speed patching workflows, and embed agentic AI safeguards directly into their development pipelines to survive an attack surface that now scales at the speed of computation. If you thought keeping up with patch Tuesday was difficult, wait until you are triaging ten thousand zero-days discovered by an AI before lunch.”
At least outside of cybersecurity circles, the expansion of Mythos AI was overshadowed in the news by Anthropic’s announcement of its public filing with an anticipated potential value of $1 trillion USD. This is in spite of there still being no clear path to profitability for the company, as it almost certainly operates at a loss of tens of billions of dollars monthly. However, it has seen very strong enterprise demand for its existing Claude Code products. Positioning for the enterprise cybersecurity market is also likely a component of the Project Glasswing expansion, as Anthropic looks to both assuage concerned regulators and court government and critical infrastructure contracts.
