OpenAI expands GPT-5.5-Cyber access to challenge Anthropic Mythos in AI-powered defense
10 Sources
[1]
OpenAI gives European companies access to its latest models to bolster resilience
LONDON, May 12 (Reuters) - U.S. artificial intelligence giant OpenAI said it was granting access to its latest models including GPT-5.5-Cyber to Deutsche Telekom, BBVA and dozens more European companies to help bolster their resilience to vulnerabilities in their systems. Other companies added to
[2]
OpenAI rolls out new model for cybersecurity teams a month after Anthropic's Mythos debut
OpenAI CEO Sam Altman speaks during the BlackRock Infrastructure Summit on March 11, 2026, in Washington. OpenAI on Thursday announced that GPT-5.5-Cyber, a variation of its latest artificial intelligence model, is rolling out in a limited preview capacity to vetted cybersecurity teams, a month
[3]
OpenAI rolls out new model for cybersecurity teams a month after Anthropic's Mythos debut
* OpenAI introduces GPT‑5.5‑Cyber following uproar around Anthropic Mythos * It's a modest upgrade focused on permissive cybersecurity tasks like vuln triage and malware analysis * Access is limited to vetted teams in the Trusted Access for Cyber (TAC) program, unlike Anthropic's more restricted
[4]
How dangerous is Anthropic's Mythos AI? | Bruce Schneier
The system's power is comparable to others - but it still has frightening implications for the future of hacking Last month, Anthropic made a remarkable announcement about its new model, Claude Mythos Preview: it was so good at finding security vulnerabilities in software that the company would
[5]
OpenAI makes its rival to Anthropic's Mythos more widely available to cyber defenders
Why it matters: Recent security testing suggests that GPT-5.5 model is nearly as good at finding and exploiting software bugs as Anthropic's Mythos Preview. The capabilities of the new models have sparked an urgent debate in Silicon Valley and the White House about how to keep them out of the hands
[6]
The AI Wars Are Having a Surprising Cybersecurity Benefit. Here's How
The AI company announced on Thursday that it is releasing a limited preview of a model called GPT-5.5-Cyber to vetted cybersecurity professionals -- in much the same way that Anthropic announced it had released a preview of Mythos to a select group in early April, and launched a public beta of
[7]
EU says OpenAI offers to open access to cybersecurity model, Anthropic not there yet
The European Commission has expressed enthusiasm for OpenAI's generous offer to grant open access to its advanced cybersecurity capabilities, a move set to fortify security standards and enhance public safety across Europe. Conversely, rival entity Anthropic has remained silent on any similar
[8]
OpenAI Shares Its Latest Cyber Model With EU, Anthropic Reportedly Holds Back Mythos - Alphabet (NASDAQ:G
OpenAI has agreed to provide the European Union (EU) with access to its latest cyber model. However, Anthropic reportedly remains reluctant to release its Mythos to the EU. The Sam Altman-led company has decided to grant access to its GPT-5.5-Cyber model to EU partners, including businesses,
[9]
OpenAI Launches Cybersecurity Preview To Challenge Anthropic's Mythos
OpenAI introduced a preview of GPT-5.5-Cyber on Thursday to a limited number of cybersecurity defenders, as part of its "broader work to build the core infrastructure for AI." For more sensitive workflows, OpenAI said it is offering GPT‑5.5‑Cyber in a limited preview with stronger verification and
[10]
OpenAI gives European companies access to its latest models to bolster resilience
LONDON, May 12 (Reuters) - U.S. artificial intelligence giant OpenAI said it was granting access to its latest models including GPT-5.5-Cyber to Deutsche Telekom, BBVA and dozens more European companies to help bolster their resilience to vulnerabilities in their systems. OpenAI's "Trusted Access
Share
Copy Link
OpenAI has launched GPT-5.5-Cyber, granting vetted cybersecurity teams and dozens of European companies access to its latest AI model for defensive work. The move comes one month after rival Anthropic's restricted Mythos release and marks a more open approach to deploying AI for cybersecurity, with testing showing both models possess comparable capabilities in finding software vulnerabilities.
OpenAI launches GPT-5.5-Cyber with broader access strategy
OpenAI has rolled out GPT-5.5-Cyber to vetted cybersecurity teams through its Trusted Access for Cyber program, marking a strategic counter to Anthropic Mythos Preview's limited release just one month earlier
2
. The AI for cybersecurity model is specifically trained to be more permissive on security-related tasks compared to the generally available GPT-5.5, enabling cybersecurity teams to conduct vulnerability identification, triage, patch validation and malware analysis with fewer restrictions3
. According to OpenAI, the preview is not intended as a major capability leap but rather provides specialized access behavior for advanced defensive workflows that would be more challenging with standard safeguards2
.
Source: Benzinga
European companies gain access to defensive AI tools
OpenAI has granted access to its latest AI models including GPT-5.5-Cyber to Deutsche Telekom, BBVA, and dozens more European companies to bolster their resilience against system vulnerabilities
1
. Other European companies added to the scheme include Spain's Telefonica, Britain's Sophos, and German financial services firm Scalable Capital1
. Emmanuel Marill, OpenAI's MD for EMEA, emphasized the need to strike a balance between access, usefulness and safety as AI models become more capable, stating the company needs to "block dangerous activity, while making sure trusted defenders have tools that are genuinely useful in protecting systems, finding vulnerabilities and responding to threats quickly"1
. OpenAI has also offered the European Commission open access to cybersecurity features, though Brussels noted that Anthropic had not been as forthcoming1
.Competing approaches to AI model deployment
OpenAI and Anthropic are pursuing markedly different strategies for rolling out their cyber-capable generative AI systems. While Anthropic limited Claude Mythos Preview access to approximately 40 organizations through its Project Glasswing initiative, OpenAI is taking a more open approach by releasing one version with stricter guardrails publicly while creating a version with fewer safeguards for vetted companies
5
. Testing by the UK's AI Security Institute found that GPT-5.5 was able to complete a 32-step simulated corporate cyberattack in 2 out of 10 test runs, while Mythos succeeded in 3 out of 10 runs—marking the first time any AI models had successfully completed that test5
. A source familiar with the capabilities told Axios that GPT-5.5-Cyber's abilities were roughly on par with Mythos, with one major recent test putting Mythos narrowly ahead5
. The company Aisle even reproduced Anthropic's published results with smaller, cheaper models4
.
Source: TechRadar
Related Stories
Implications for offensive and defensive cybersecurity
The capabilities of these advanced AI models have sparked urgent debate in Silicon Valley and the White House about keeping them from bad actors while enabling legitimate defensive cybersecurity work
5
. Security expert Bruce Schneier notes that while attackers will use these capabilities to find and automatically hack software vulnerabilities in critical systems worldwide, defenders will simultaneously use them to find and patch those same system vulnerabilities4
. Mozilla used Mythos to find 271 vulnerabilities in Firefox, which have since been fixed and will never again be available to attackers4
. However, Schneier warns of a more dangerous short-term future, as many systems aren't patchable and finding and exploiting appears easier than finding and fixing4
. The White House is actively discussing executive actions that could change how the federal government is involved in future model rollouts, with Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent having met with major US bank CEOs to discuss Mythos last month2
. OpenAI is scaling its Trusted Access for Cyber program to thousands of verified individual defenders and hundreds of teams responsible for defending critical software, while defenders will still be blocked from certain tasks like credential theft and writing malware5
. Former British finance minister George Osborne, who heads OpenAI's "OpenAI for Countries" initiative, sent an explanatory letter to the European Commission arguing that democratizing access to defensive tools could strengthen shared security and support public safety1
. Organizations will need to adapt their security to this new reality as bug hunting and responsible deployment of these powerful AI models reshape the cybersecurity landscape.
Source: Inc.
References
Summarized by
Navi
[3]
[4]
Related Stories
OpenAI warns upcoming AI models will likely pose high cybersecurity risk with zero-day exploits
11 Dec 2025•Policy and Regulation
OpenAI halts Astra AI model development over critical cybersecurity threshold concerns
07 Aug 2026•Technology
OpenAI briefs government agencies and Five Eyes on GPT-5.4-Cyber as AI cybersecurity race heats up
22 Apr 2026•Technology
Recent Highlights
1
AI Hallucination Nearly Sparked US Military Confrontation With Chinese Ship in Middle East
Technology
2
OpenAI's AI Solves Navier-Stokes Problem But Ignites Existential Crisis in Mathematics Community
Science and Research
3
Anthropic's Claude Now Leads 26% of AI Development Work, Company Reveals New Transparency Metrics
Technology
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
