At this month’s NATO summit in Ankara, allies announced billions of dollars in new arms deals and reaffirmed their commitment to spend more on defense. European governments have made the pledge, and the money is real: European defense spending has doubled since 2019, and by 2030, European NATO member countries are projected to spend in excess of €800 billion a year, up €300 billion from 2025, with equipment spending alone nearly doubling.

But committing money is the easy part. The harder question is whether Europe’s defense industry can turn it into equipment fast enough to matter. Europe’s largest defense manufacturers’ order books now average more than five years for production, and some are closer to nine. Money is flowing in faster than industry can turn it into equipment. But a purchase order is not equipment that can be deployed on the ground and the air.

European countries have long duplicated capabilities rather than pooling them.

The bottleneck sits in the defense industrial system. Deterrence relies on the chain from funding to contracts, then through production, deployment into services, then rapid innovation in the field. Europe’s next goal comes after the spending promise. The continent fields six times as many weapons platforms as the United States, because countries have long duplicated capabilities rather than pooling them. Production ends up split across many small runs that never reach an efficient scale. Ukraine, under pressure, has shown how fast a defense system can move, adapting tactics in weeks and building drone detection networks from consumer electronics. Europe needs to catch up and then accelerate.

Four moves would help Europe accelerate.

The first is multi-speed procurement. Software-led systems such as drones and targeting improve in rapid cycles throughout their deployment and need procurement that can keep up. Israel’s Iron Dome started out as far less capable than it is today and improved continuously in service. European defense ministries have already set up high-speed procurement units with dedicated teams and greater risk tolerance. These need to become mainstream, rather than the exception.

Collaboration in procurement, maintenance and training brings costs down and delivery forward.

The second is military collaboration to reduce fragmentation. Collaboration in procurement, maintenance and training brings costs down and delivery forward. The Tempest project, where the U.K., Italy and Japan are jointly building a next-generation fighter, demonstrates the model: shared development costs that no single country could carry alone. Recent bilateral maritime agreements, and Romania’s use of EU funding to buy European while expanding production at home, show the same logic spreading.

The third is industrial consolidation, which is already underway and needs to move faster. Companies are driving it themselves. Airbus, Leonardo and Thales have agreed to merge their space divisions into a single joint venture with roughly €6.5 billion in revenue and 25,000 employees, and European defense mergers and acquisitions rose 35 percent year over year in the first half of 2025. McKinsey analysis finds that consolidation across key supply chain segments could unlock around €9 billion in annual synergies, more than the current equipment budgets of 24 of Europe’s 30 NATO members. The deepest opportunity sits below the big primes, among the thousands of tier two, three and four suppliers that still duplicate one another’s work. Europe can speed this up by harmonizing requirements, reducing national carve-outs and letting industry do the combining. Consolidation is only half the task. Europe also needs to build sheer capacity — more shipyards, more assembly lines, more of the physical plants that turn orders into hardware — and the capital to fund it. In several categories, Europe simply lacks enough places to build.

Real deterrence means difficult choices, and a public that understands the importance and the cost of security.

The fourth is regulatory unlocking. Full scale-up demands skilled workers retrained, accredited and security cleared from other industries; production sites with preapproved permitting; and alignment of export controls across European allies. These regulatory unlocks now need the same energy and focus as the funding commitment debate. 

Real deterrence means difficult choices, and a public that understands the importance and the cost of security. That conversation is only beginning in much of Europe. It must include the potential for “gray zone” cyber strikes on hospitals, arson at industrial sites, drones disrupting ports, undersea data cables cut — these have all occurred, but many citizens do not yet recognize this as having malicious intent.

The opportunity in getting it right is significant. McKinsey and GLOBSEC estimates indicate that every euro of spending on European-manufactured equipment generates two euros of revenue across the European supply chain, and an additional €165 billion a year in equipment spending could create up to 1.2 million jobs. The coming years will reveal how effectively Europe is able to scale up to protect its territory and citizens, and how much of the promised investment becomes lasting deterrence and European jobs. Getting there depends on the whole ecosystem — governments, industry and investors — moving together. Increased spending is important. Spending it effectively matters more.

Jonathan Dimson is a senior partner in McKinsey’s London office. Mikael Robertson is a senior partner in the Stockholm office.

Europe wants to kick its U.S. tech habit, but it’s struggling to detangle itself from American data analytics firm Palantir. 

The problem is, the company’s tools are already deeply embedded in critical sectors across the bloc like policing, national defense and health systems. A report released today also shows that the tech firm has found ways to pay minimal taxes in Europe. On the show, hosts Zoya Sheftalovich and Ian Wishart discuss European alternatives and why detaching is easier said than done.

Next, our colleague Seb Starcevic has interviewed Mediterranean Commissioner Dubravka Šuica. The role that was originally thought to be low profile has instead thrust the Croatian official into the limelight as various crises hit the Middle East and the Mediterranean.

Plus: Three new foods are being added to the EU’s list of products with geographic indicators. These are the labels given to products that can only be made in a specific region, like Champagne or … yes, you guessed it, halloumi.

Questions? Comments? Get in touch! You can message us or send a voice message to our WhatsApp here or at +32 491 05 06 29.

Leading artificial intelligence models from Anthropic and OpenAI created fake online personas and tried to deceive human coders into abetting a cyberattack during a recent safety evaluation, the U.K.’s AI Safety and Security Institute disclosed Tuesday.

It marks the latest case in which a powerful AI system has attempted a digital attack on an unwitting third party without direct prompting during such an evaluation — heightening concerns the powerful technology is advancing too fast for responsible oversight.

The disclosure is likely to ignite fresh calls in Washington and Silicon Valley for more rigorous regulation of the AI industry, particularly over frontier models with advanced capabilities to detect and launch cyberattacks. It comes just days after similar testing mishaps involving some of the same models from OpenAI and Anthropic sparked urgent calls for new AI safety regulation and a push within Silicon Valley to slow the rapid pace of AI development.

Like its U.S. counterpart, AISI routinely conducts security evaluations to better understand what dangers both new and soon-to-be-released AI models pose to public health and safety. But even the digital security body said the actions it uncovered by Anthropic’s Claude Mythos 5 and ChatGPT 5.6 — the latest publicly released model from either AI lab — were unlike anything it had seen before.

“This is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world,” AISI said in a 35-page technical report accompanying a blog post Tuesday.

AISI said the malicious activity began on July 25. The AI safety evaluator detected it and launched an investigation on July 28, when it spotted “unusual data transfers” stemming from a cyber evaluation it was conducting on Mythos 5 and ChatGPT 5.6.

Overall, AISI determined that Mythos 5 and ChatGPT 5.6 took “autonomous, unsanctioned action on the live internet, targeting real people and organizations” during 10 of the 122 evaluations the research body conducted and isolated for review. The majority of those actions, it said, involved Mythos 5 and arose during a “single, sustained line of activity.”

In the most alarming incident it uncovered, Mythos 5 attempted a supply chain attack — a sophisticated technique often deployed by North Korean and Russian cyber operatives, in which hackers bury malware inside legitimate code.

To do so, Anthropic’s model created “multiple fake identities” on developer platform GitHub and used them to send messages “pressuring” an open-source software engineer to unwittingly introduce a bugged update into code widely available on the popular site, AISI said. When that effort failed, the AI “edited its earlier activity to appear harmless” and “considered adopting a fresh identity to continue,” AISI added, a sign the model was intent on repeating the ruse.

As part of the same effort, Mythos 5 also sent direct messages over GitHub to software engineers that contained malware.

In another sign of deceitful behavior AISI uncovered in its investigation, multiple AI agents it was testing appeared to communicate with one another about how to convince real engineers using GitHub to trust them. “One agent left public messages on GitHub offering collaboration with other agents working on the same challenge,” AISI wrote.

AISI’s blog and technical assessment make no mention of whether the models also attempted to exploit previously unknown software bugs — called zero-days — during the evaluation.

Last month, OpenAI disclosed that GPT 5.6 and another of its models escaped onto the open internet during what was supposed to be a controlled test, and then hacked another company in a first-of-its-kind, autonomous breach.

In response, Anthropic launched an investigation into whether any of its models took illicit action during recent testing and discovered Mythos 5 and two other models had hacked three organizations during tests dating back to April.

In a statement, an Anthropic spokesperson said they are “grateful” to AISI for their leadership and that this review underscores the need for “a broader conversation about how to safely evaluate increasingly capable AI agents.”

The spokesperson added: “As we shared after disclosing our own incident last week, the field needs stronger, shared standards for how evaluation environments are built and secured. We look forward to partnering with the UK AISI to learn more about this incident as we conduct our own investigation.”

An OpenAI spokesperson referred POLITICO to a blog post about the incident that went up Tuesday evening. “We are committed to working across the industry to strengthen shared practices for conducting high-risk evaluations safely, including convening stakeholders such as national AI institutes, independent evaluators, other AI labs, and other groups in the coming weeks,” the blog read.

AISI stressed in its blog that the malicious activity it disclosed Tuesday took place under “deliberately permissive conditions” so they could assess the safety risks posed by the two models. This included granting the models access to the internet, unlike the earlier incidents detailed by Anthropic and OpenAI.

AISI also noted the models were intentionally stripped of internal guardrails that block malicious behavior. AISI was only able to disable those controls because of its role testing Mythos 5 and ChatGPT 5.6.

Still, AISI said the incidents highlighted the need for greater monitoring of model behavior during testing, and tighter controls over their access to the internet.

The Trump administration is finalizing a voluntary framework under which AI labs would submit powerful models they want to release to the public for federal safety testing. But it has not yet made the framework public, and it includes no provisions for models AI labs are developing internally.

The incidents last month from OpenAI and Anthropic both involved models not intended for public release.

Some cyber experts say recent incidents highlight deeper questions around AI development, such as who is liable when AI systems break federal hacking laws.

“If any of these were human-originated, they would lead to clear and vigorous prosecution. I think it’s time for a serious discussion about updates to existing computer security law,” said Marc Rogers, a hacker and prominent cybersecurity expert.