The choice is coming
Anthropic lost its Pentagon contract last week after refusing to remove safeguards that blocked its AI from being used for domestic mass surveillance or fully autonomous weapons. President Trump called the company's staff "leftwing nut jobs" and directed federal agencies to phase out their technology.
OpenAI, meanwhile, has no such restrictions. The company revised its usage policy in January to allow military applications, explicitly permitting work with defence and intelligence agencies.
For Australian AI startups, this is not a distant debate. It is a preview of decisions they will make in the next 12 to 24 months.
Australia is pushing dual-use tech
The federal government is actively funding AI development with defence applications. The Advanced Strategic Capabilities Accelerator (ASCA) is backing projects that bridge commercial and military use. Defense Connect reported in February that ASCA is prioritising AI for intelligence analysis, autonomous systems, and decision support.
That funding comes with expectations. If your startup takes government money to build AI that analyses satellite imagery or processes intelligence data, you are building dual-use technology. The line between commercial and military deployment gets thin fast.
What this means for sales teams
If you are selling AI tooling, this matters in three ways:
1. Customer concentration risk just got real. If 30% of your ARR comes from government contracts and those contracts require removing ethical safeguards, your board will have opinions. So will your team. Anthropic just proved you can lose major revenue by holding a line.
2. Compliance is about to get complicated. Enterprise buyers are already asking about data sovereignty and model governance. Add military use cases and you are navigating export controls, security clearances, and audit requirements that most sales teams are not staffed to handle.
3. Talent retention becomes harder. Anthropic's stance cost them Pentagon money, but it probably helped them retain AI researchers who care about safety. If your company takes the opposite path, expect attrition from technical teams. That affects your product roadmap, which affects what you can sell.
The numbers
Anthropic was one of three preferred Pentagon AI providers. That contract was worth an estimated $500 million across multiple agencies. OpenAI and another provider (unnamed in reports) remain approved.
In Australia, ASCA's budget sits at $3.4 billion over four years. A meaningful chunk of that will flow to AI startups building defence-adjacent technology. For early-stage companies, that is survival capital. It is also capital with strings.
No easy answers
This is not a take about what startups should do. It is a note that the choice is no longer hypothetical. Australian AI companies will get offers: government funding, enterprise contracts, revenue that changes your burn rate overnight.
Some will take it. Some will not. Both decisions have consequences for pipeline, team stability, and long-term positioning.
Worth noting: the Anthropic situation is still developing. The company has not publicly confirmed whether it will revise its terms or walk away from government revenue entirely. That decision will set a precedent either way.
For sales leaders: if your product touches AI, start asking your exec team where the company stands on dual-use applications. Better to have that conversation now than when a $2 million government deal is sitting in your pipeline with a clause your engineers will not sign off on.