Economy May 12, 2026 07:08 AM

OpenAI Opens Latest Models to European Firms Under Defensive Access Program

Access to GPT-5.5-Cyber and other models is being extended to verified organisations in key sectors to help shore up resilience against growing AI-enabled cyber risks

By Ajmal Hussain

OpenAI is granting select European companies access to its newest models, including GPT-5.5-Cyber, through a program aimed at equipping trusted defenders with tools for identifying and responding to vulnerabilities. The initiative covers firms across finance, telecommunications, energy and public services, and coincides with broader company moves to supply cybersecurity features to European regulators and to scale an enterprise AI unit backed by over $4 billion in initial capital.

OpenAI Opens Latest Models to European Firms Under Defensive Access Program

Key Points

  • OpenAI is granting verified European firms, including Deutsche Telekom and BBVA, access to its newest models such as GPT-5.5-Cyber through the "Trusted Access for Cyber" programme.
  • The programme targets critical sectors - financial services, telecoms, energy and public services - and provides specific safeguards for defensive cybersecurity work.
  • OpenAI has offered the European Commission open access to its cybersecurity features; the company is also creating a new enterprise AI unit with over $4 billion in initial investment and will acquire AI consulting firm Tomoro to scale the unit quickly.

OpenAI has begun providing verified European companies with access to its latest AI models, including GPT-5.5-Cyber, as part of an effort to strengthen corporate resilience to vulnerabilities in their systems. The program, described by the company as "Trusted Access for Cyber," targets organisations in critical sectors such as financial services, telecoms, energy and public services.

The initiative will make advanced models available to a range of firms, including Deutsche Telekom and BBVA, along with dozens of other European companies. Access under the programme includes specific safeguards intended for defensive cybersecurity work, OpenAI said.

OpenAI's managing director for Europe, the Middle East and Africa, Emmanuel Marill, framed the approach as a balance between access, usefulness and safety as AI systems increase in capability. "We need 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," he said on Tuesday.

The move follows concerns raised by the arrival of competing frontier models. The release of Mythos by Anthropic last month has been cited as raising the stakes for firms such as banks and other companies, because such models can perform high-level coding tasks that enable them to identify cybersecurity weaknesses and, potentially, ways to exploit them.

OpenAI has offered the European Commission open access to its cybersecurity features, according to statements attributed to Brussels, while the commission has said Anthropic has not been as forthcoming on similar terms. Within OpenAI, the "OpenAI for Countries" initiative is being led by former British finance minister George Osborne, who on Monday sent an explanatory letter to the Commission arguing that broader access to defensive tools could bolster shared security, support public safety and align with European priorities.

In parallel with the trusted-access program, OpenAI said on Monday it was creating a new company with more than $4 billion in initial investment to help organisations build and deploy AI systems, and that it would acquire AI consulting firm Tomoro to accelerate the new unit's scale-up.


Context and implications

OpenAI's programme is designed to equip vetted organisations with advanced model capabilities and tailored safeguards so they can perform defensive tasks - from vulnerability discovery to incident response - without enabling misuse. The arrangement explicitly focuses on trusted entities in sectors where failures could have systemic effects.

Risks

  • Advanced frontier models like Anthropic's Mythos can code at a high level, increasing the potential for both identifying and exploiting cybersecurity vulnerabilities - a risk the programme seeks to mitigate. (Impacted sectors: banking, corporate IT, critical infrastructure)
  • Differences in cooperation with European authorities among AI firms could create asymmetries in defensive capabilities and regulatory transparency. (Impacted sectors: public services, regulators, financial services)
  • Concentrating advanced defensive tools with a subset of verified organisations introduces questions about access and potential gaps for entities not included in the programme. (Impacted sectors: small and mid-sized enterprises, non-participating public bodies)

More from Economy

Trump Says He Will Discuss Iran with Xi in China Visit, But Says He Doesn’t Need Beijing’s Help May 12, 2026 U.S. Posts Reduced $215 Billion April Surplus as Refunds and Outlays Rise May 12, 2026 Bundesbank's Nagel Says Iran Conflict Could Force ECB to Raise Rates May 12, 2026 EIA Sees U.S. Electricity Use Climbing to New Highs Through 2027 May 12, 2026 Chicago Fed’s Goolsbee Flags Broadening Inflation, Cites Services as Main Concern May 12, 2026