OpenAI and the Ukrainian government have agreed to a partnership that will give the country's cyber defenders access to frontier AI models and subsidized computing resources to protect critical infrastructure from relentless cyberattacks, CyberScoop reported Wednesday. The agreement, announced at OpenAI's New York office, runs through the company's Daybreak program and includes a pledge of more than $1 billion in subsidized tokens to support the initiative.
The deal lands at a moment when Kyiv's digital defenders are contending with one of the most intense cyber threat environments on the planet — and as OpenAI works to position its models as essential tooling for governments racing to harden their networks against attackers who are adopting the same technology. For more context on this story, see our ongoing artificial intelligence updates.
Machine-Speed Defense for Critical Systems
Dmytro Kushneruk, Ukraine's consul general in San Francisco, outlined during a panel discussion how the tools will be deployed: incident response, threat triaging, login analysis, inventorying systems, code analysis and validating vulnerabilities. In nearly all of those cases, he said, the primary benefit is executing the work at machine speed — a force multiplier for a defensive workforce stretched thin by war.
Incident response illustrates the gap. Human analysts must review, in Kushneruk's words, "thousands and thousands of these logs and they have to find what's really important, that's why AI can give capable defenders really much greater advantage and leverage."
He was careful to frame the technology as an accelerant for skilled people rather than a substitute for them. "This is why the object is not to replace the cyber defender with AI, but to make sure the cyber defender acts faster," Kushneruk said. For a country whose electricity and water systems have been under sustained attack for years, he added, the partnership "is really not about protecting computers, it is about actually keeping our country running."
A Country Facing 6,000 Attacks a Year
The scale of the problem explains the urgency. Ukraine faces approximately 6,000 cyberattacks per year — roughly 15 per day, according to Kushneruk — layered on top of physical strikes against the same infrastructure.
The country has been preparing for this since long before Russia's 2022 full-scale invasion. Russian military hackers shut down portions of Ukraine's power grid in 2015, an attack that became a global wake-up call about the vulnerability of civilian infrastructure. Kushneruk acknowledged that Ukraine was "maybe not so much prepared" for the fallout in 2015, but said its defenses matured steadily — resilience the world saw again in 2025, when Ukrainian trains kept running after hackers struck the country's railway system.
That track record has turned Ukraine into something of a living laboratory for critical infrastructure defense. Some U.S. national security experts and congressional committees have explicitly cited it as a model for American industry.
What Daybreak Actually Is
Daybreak is OpenAI's program for giving vetted private organizations and governments access to unreleased frontier models for defensive cybersecurity work. The company expanded it in August with dedicated "Red" and "Blue" tracks for offensive testing and defensive workflows, a specialized model variant, and partnerships with 16 major cybersecurity vendors.
The Ukraine partnership is the program's most high-profile government commitment to date, and the $1 billion subsidized-token pledge is substantial even by the standards of an industry that has grown accustomed to nine-figure announcements. Subsidized tokens matter in practice: running frontier models continuously against log streams, code repositories and network telemetry across an entire nation's infrastructure is compute-intensive, and cost is a real constraint for a wartime economy.
A Signal for U.S. Critical Infrastructure
The announcement was also aimed, implicitly, at Washington. Naz Durakoğlu, minority staff director of the U.S. Senate Foreign Relations Committee, said there is "pretty much across the board" agreement between the parties on adopting defensive AI tools for U.S. critical infrastructure operators, though regulatory questions remain sticking points.
„Es besteht ein breites Verständnis dafür, dass dies ein großes Problem ist, und ich möchte sagen, dass der Anblick dessen, was die Ukraine tagtäglich durchmachen muss, auch ein großer Weckruf für unsere Mitglieder auf parteiübergreifender Basis ist“, sagte Durakoğlu.
OpenAI hat sich das ganze Jahr über auf diesen Markt konzentriert. Firmenpräsident Greg Brockman unterzeichnete einen Anfang 2026 veröffentlichten offenen Brief, in dem er zu kollektiven Maßnahmen aufrief, um mithilfe von KI-Modellen Schwachstellen zu finden und zu beheben, bevor Gegner – darunter ausländische Regierungen und Cyberkriminelle – Zugang zu denselben Fähigkeiten erhalten. Laut Politico traf sich CEO Sam Altman im Juli mit US-Energieunternehmen, um den Einsatz von KI zum Schutz des amerikanischen Stromnetzes zu besprechen.
Das Dringlichkeitsargument
Sasha Baker, Leiterin der nationalen Sicherheitspolitik bei OpenAI, sagte, das Unternehmen halte es für „dringend“, ähnliche Vereinbarungen mit anderen Regierungen und Branchen zu schließen. Die Argumentation ist ein Race-Condition-Argument: Mit zunehmender Verbreitung von KI-Fähigkeiten werden sich auch offensive Akteure diese aneignen, sodass es ein sich schließendes Zeitfenster gibt, in dem Verteidiger einen Vorteil aufbauen können.
„Wir haben dieses Zeitfenster, um Maßnahmen zu ergreifen, und wir sind wirklich motiviert von der Idee, dass wir dringend handeln müssen“, sagte Baker.
Diese Formulierung wird jedem bekannt sein, der die Debatte über KI und Cybersicherheit verfolgt hat, wo dieselben Modelle, die die Schwachstellenerkennung für Verteidiger automatisieren, andererseits Angriffe beschleunigen können. Die Ukraine – ein Land, das mehr feindliche Cyberaktivitäten als jedes andere Land absorbiert hat – ist jetzt der deutlichste reale Test dafür, ob Grenz-KI das Gleichgewicht zugunsten der Verteidiger verschieben kann.
---
Der KI einen Schritt voraus seinErhalten Sie die neuesten KI-Nachrichten, Analysen und Durchbrüche – alles an einem Ort.
Weitere KI-Neuigkeiten lesen →
