Advertise with Pune MediaAdvertise with Pune MediaAdvertise with Pune MediaAdvertise with Pune Media

Top 5 This Week

Advertise with Pune MediaAdvertise with Pune MediaAdvertise with Pune MediaAdvertise with Pune Media

Related Posts

300 Billion Tokens: How NATO Is Using OpenAI’s Hacking AI

Editorial Disclosure: This article is curated from reporting by the original publisher credited below. It was selected and published automatically under the Pune.Media Editorial Policy and is not original Pune.Media reporting.

Original Coverage & Source Attribution: www.trendingtopics.eu

Set Trending Topics as a preferred source on Google.

NATO now has A.I. standing guard, too: OpenAI has just disclosed that the military alliance is using its cybersecurity models to protect Allied networks, and on a large scale. NATO has used up to 300 billion input and output tokens a month as part of the collaboration, according to a blog post by OpenAI.

What OpenAI Does for NATO

OpenAI’s partner on the NATO side is the NATO Communications and Information Agency, or NCIA, the alliance’s IT and communications agency. Through it, NATO’s core institutions get access to OpenAI’s advanced cybersecurity models. They are used through Daybreak, OpenAI’s program that puts A.I. in the hands of trusted defenders for authorized security work. According to OpenAI, that ranges from reviewing software and investigating suspicious activity to validating vulnerabilities and developing and testing fixes.

NATO is also meant to benefit in terms of skills. The collaboration strengthens the cyber expertise of the multinational teams that protect Allied networks around the clock, OpenAI wrote, and the lessons learned are meant to benefit other defenders in Europe and across the alliance.

What 300 Billion Tokens Cost

Neither OpenAI nor the alliance has disclosed what NATO pays. A rough calculation using the API list prices published by the comparison platform BenchLM shows the order of magnitude, however. Assuming 90 percent of the tokens are input and 10 percent output, the monthly cost would be:

Model Price per million tokens (input/output) Cost per month
GPT-6 Cyber $20/$100 about $8.4 million (7.5 million euros)
GPT-6 Astra $10/$50 about $4.2 million (3.8 million euros)
GPT-6.1 Sol $2/$10 about $0.84 million (0.75 million euros)

If everything ran on the specialized GPT-6 Cyber, the annual total would come to about $100 million (about 90 million euros). The real bill is likely to be considerably lower: The 300 billion is a peak figure, cached input costs only a tenth of the regular price, and OpenAI offers Daybreak access on subsidized terms anyway.

A.I. Agents at Crossed Swords

Daybreak was also used at the Crossed Swords 2026 exercise, organized by the Tallinn-based NATO Cooperative Cyber Defence Centre of Excellence, or CCDCOE. New this year was a dedicated training stream in which participants built, deployed and evaluated A.I. agents for cyber operations. The humans worked as operators alongside the agents, with an emphasis on maintaining human oversight, according to OpenAI. The company said it remains fully committed to NATO’s principles for responsible A.I. adoption.

Daybreak Expands in Europe

NATO is not the only government customer. OpenAI launched Daybreak in September and committed $1 billion (about 900 million euros) in subsidized access over six months. The program has two tiers: Blue, with standard models for routine work, and Red, with specialized cyber models for more sensitive tasks, including the new GPT-6 Cyber. Ukraine gets Daybreak for free to protect civilian infrastructure such as hospitals and power plants.

European agencies are also reporting early results, as summarized by StartupHub.ai:

  • ENISA: The E.U.’s cybersecurity agency found four vulnerabilities in software used across E.U. institutions, all of which have been fixed.
  • CERT-EU: In a joint review of 27 projects with ENISA, 95 percent of 88 potential issues were confirmed.
  • CERT Polska: Poland’s computer emergency response team helped find six vulnerabilities in MikroTik’s RouterOS router operating system.

National agencies in France, Germany and Poland are also involved.

The Race for Military A.I.

OpenAI is not the only U.S. tech company that has discovered the defense business. NATO already uses Palantir’s Maven Smart System, which analyzes data from drones and satellites for military targeting. Google has agreed with the Pentagon that the U.S. military may use its A.I. models for any lawful government purpose, which has led to resignations in protest at Google DeepMind. OpenAI and xAI also have Pentagon contracts. Anthropic, by contrast, was blacklisted by the U.S. Defense Department after refusing to remove safeguards against the use of its models for weapons and surveillance.

Dependent on Silicon Valley?

There are still critical questions. NATO must first learn how many tokens it consumes, what its usage patterns look like and how model economics work in order to control costs and compare vendors, Dylan Browne, the NCIA’s general manager, cautioned, according to StartupHub.ai. And about 5 percent of the issues flagged by the A.I. were not confirmed by the software maintainers.

The broader debate concerns Europe’s dependence on U.S. technology. Since President Trump has repeatedly threatened to leave NATO, concerns about a kill switch in Washington have grown. The Munich defense start-up Helsing is therefore positioning its military A.I. as “from Europe for Europe” against Palantir. And control over the models themselves is also an issue: Just on Monday, California sued OpenAI because the company’s experimental A.I. agents allegedly escaped their test environment and attacked Hugging Face, among others.

Popular Articles