The biggest enterprise trust signal in AI history
The Pentagon formalised agreements today with seven leading AI companies, clearing them to deploy capabilities on Defense Department classified networks at Impact Levels 6 and 7, the highest tiers for sensitive military data. The companies are SpaceX, OpenAI, Google, NVIDIA, Reflection, Microsoft, and Amazon Web Services. Oracle joined the same day.
The Pentagon described the move as accelerating ‘the transformation toward establishing the United States military as an AI-first fighting force.’ This is not a pilot. This is operational integration at the core of US command and control.
For every CIO and board risk committee in the world, Impact Level 6/7 clearance is the strongest possible enterprise trust signal. These platforms have been stress-tested against the most demanding security requirements on earth.
One company failed the test
Anthropic was shut out. The company had been locked in a dispute with the Pentagon over guardrails for military AI use, and the Defense Department labelled it a supply-chain risk, a designation normally reserved for companies from adversary nations.
In February 2026, Defense Secretary Pete Hegseth issued an ultimatum: ‘Effective immediately, no contractor, supplier, or partner that does business with the United States military may conduct any commercial activity with Anthropic.’
OpenAI navigated the same ethical tension differently. CEO Sam Altman announced a deal that included ‘technical safeguards’, including ‘prohibitions on domestic mass surveillance and human responsibility for the use of force.’ The Pentagon accepted those terms. Anthropic’s were rejected.
The lesson for any business dealing with a dominant institutional buyer: compliance with operational requirements is not optional, regardless of how principled the refusal.
This confirms a structural shift, not a one-off
The May agreements build on a pattern. In March 2026, Deputy Secretary of Defense Steve Feinberg designated Palantir’s Maven Smart System a formal programme of record, locking in long-term funding for its weapons-targeting AI. Palantir’s stock has doubled in the past year, pushing its market capitalisation to nearly US$360 billion. Its US Army contract carries a ceiling of up to US$10 billion.
Fortune’s analysis identified this period as a structural inflection point: tech defence startups moving from ‘dabbling in limited pilot projects’ to ‘core missions with the kind of fixed-priced deals that have long been standard among established defence contractors.’
AI has moved from speculative promise to mission-critical procurement. The market will price that accordingly.
Why this matters in Auckland, not just Arlington
The seven companies the Pentagon just anointed are not abstract US defence names. They are the backbone of how New Zealand businesses operate today.
According to the AI Forum’s March 2025 survey, 82% of NZ businesses reported some level of AI use. Of those, 72% were using off-the-shelf tools: ChatGPT, Microsoft Copilot, Google Gemini. Three of the seven companies just cleared at the highest classification level. AWS underpins a significant share of NZ cloud infrastructure. NVIDIA’s chips power the models those businesses run.
New Zealand’s national AI strategy, published in mid-2025, explicitly positioned the country as an AI adopter rather than developer, with industry research suggesting generative AI alone could add NZ$76 billion to the economy by 2038. That strategy’s logic depends entirely on NZ businesses plugging into globally dominant AI platforms. Those platforms are now confirmed.
Domestically, Inland Revenue’s 2026 information release documenting its AI deployment shows NZ government agencies are already embedding commercial AI in sensitive operational systems, a parallel to the Pentagon’s move at a smaller scale.
The concentration risk nobody wants to discuss
Here is the uncomfortable truth. The Pentagon has just done the due diligence that most enterprise procurement teams can only dream of, and published the results. Seven vendors cleared. One failed. For NZ businesses already using these tools, the deals are confirmation they backed the right vendors. For those still evaluating, the shortlist just got dramatically shorter.
But that shortlist deepening NZ’s dependency on a handful of US platforms is a risk the draft national AI strategy acknowledged but did little to address. It says nothing about what happens if a vendor falls out of favour with Washington the way Anthropic did.
With 72% of NZ businesses already running on these platforms, and the Pentagon’s imprimatur now pushing more organisations toward the same vendors, the concentration is only growing. That is simultaneously the safest bet and the biggest single point of failure in New Zealand’s AI future.
Sources
- OpenAI strikes Pentagon deal with ‘safeguards’, as Trump dumps Anthropic (2026-02-28)
- Pentagon to adopt Palantir AI’s core US military system, memo says (2026-03-09)
- A turning point at the Pentagon: Anduril’s new mega-deal rewrites the rules for Silicon Valley (2026-03-22)
- AI in Action: Exploring the Impact of Artificial Intelligence on New Zealand’s Productivity – Report 2 (2025-03)
- New Zealand’s Strategy for Artificial Intelligence – Draft National Strategy (2025-05)
- Inland Revenue’s use of artificial intelligence – Information release (2026-03)