The Story

OpenAI just made a choice.

They signed a classified contract with the US Department of Defense. Their tools will now power military AI systems. Classified systems. Systems you won't be able to see or audit.

Anthropic had the same opportunity. They said no.

Not because OpenAI is wrong. But because they weighed the same problem differently.

This is what happens when two companies face the same fork in the road and walk in different directions.

What Matters Here

This isn't a technology question. It's a values question.

If you use OpenAI tools, those tools now carry a choice you didn't make. Your vendor's priorities are now entangled with yours. That matters.

For faith-led organisations, it's a theology problem. For nonprofits, it's a mission problem. For enterprises, it's a governance problem.

Most leaders don't know they have a choice. But you do.

Section A: For Faith-Led Organisations

Your Values Are Already Embedded in Your Vendor Choices

You probably don't think about this. Your board didn't approve ChatGPT. It just showed up. Someone tried it, liked it, shared it. Now it's everywhere.

But here's the thing: every tool you use is a choice about what you're willing to be part of.

OpenAI's Pentagon contract means their AI now shapes military decisions. You don't get to see those decisions. You can't audit them. You can't say ‘that's not who we are.’

Because if your tools power those systems, that is part of who you are — whether you intended it or not.

1. Just War & Weapons Technology

Your theology probably has something to say about this.

Catholic social teaching permits defensive military action, but only under strict conditions: legitimate authority, proportionality, and last resort. Protestant traditions split — some permit defence, others (Mennonite, Quaker) reject military involvement entirely. Jewish tradition distinguishes defensive operations from offensive expansion. Islamic theology allows defensive jihad but constrains offensive warfare.

What they all have in common: they require human accountability at every decision point.

Classified military AI removes human accountability. It replaces oversight with secrecy.

That's the problem.

The question: Does your tradition permit tools in military systems where you can't verify proportionality or legitimate authority?

2. Human Dignity & Autonomy

Many faith traditions say this: humans are made in God's image. That means something. It means human judgement in life-and-death decisions isn't optional. It's sacred.

It means autonomy — the freedom to choose — is fundamental.

When you use tools that delegate critical decisions to machines in classified systems, you're saying something about what humans are. You're saying their judgement can be outsourced to systems they'll never see or understand.

That's not just a technical choice. It's a theological one.

The question: Does choosing a vendor that removes human oversight align with your belief about what humans are?

3. Governance & Prophetic Witness

Faith organisations have a job: to be a prophetic voice. To call systems toward justice. To advocate for the vulnerable.

You can't advocate for people if you don't know what your tools are doing. You can't be prophetic about something you won't look at.

OpenAI's classified contracts are intentionally opaque. You're not supposed to know. You can't speak into them. You can't demand accountability.

That's complicity through silence. It's the opposite of prophetic witness.

Anthropic's approach is different. They publish their safety research. You can read it. Critique it. Hold them accountable. That's the difference between transparency and secrecy.

The question: Can you maintain prophetic integrity using tools you can't audit?

What To Do About It

Action Steps

Step 1: Define Your Position. Work with your theological leadership. Figure out what your tradition actually says about military AI. Write it down. Make it public. That's your frame.

Step 2: Audit Your Vendors. For every AI tool you use, document: military contracts? Classified work? Safety research public? Does it align with your theology?

Step 3: Make Intentional Choices. Replace OpenAI with Anthropic for values-sensitive work. Use open-source for internal work. Reserve OpenAI for commodity tasks.

Step 4: Tell Your Community. ‘We've evaluated our vendors against our values. Here's what we use and why.’ That's not purity. That's integrity.

Section B: For Nonprofits & Charities

Your Mission Depends on This

Nonprofits exist to solve a problem. Climate. Human rights. Education. Health. Pick one.

Now ask yourself: would your community support your work if they knew your tools funded military systems?

Some yes, some no. But that's the point — you don't get to decide for them. They get to decide whether they trust you with that choice.

1. Mission Alignment & Reputational Risk

A human rights organisation using OpenAI faces a contradiction. You're fighting for accountability while using tools deployed in systems where accountability is impossible.

An education nonprofit has a different problem: parents trust you with their children's futures. They don't expect your AI tools to have military genealogy.

A climate nonprofit has a third problem: you might be funding the AI systems that model climate scenarios for military purposes.

None of these is your intention. But intention doesn't matter if the structure contradicts your mission.

The question: Does your tool choice align with what your community actually believes about you?

2. Donor Requirements (This Is Changing Fast)

Major foundations are adding ‘vendor ethics’ clauses to grants. It's not universal yet. It's coming.

Some now require: ‘Use only vendors with published safety standards.’ OpenAI's classified work fails that test.

Others require: ‘Disclose all third-party data processing.’ Classified contracts can't be disclosed.

Others require: ‘Ensure human rights in vendor selection.’ Hard to verify with classified partners.

Warning

Check your grants. Seriously. You might already be in violation without knowing it.

3. Operational Resilience

If you're deeply embedded in OpenAI (API, workflows, training data), switching costs are high. But forced migration also disrupts operations.

The question: Could you migrate to Anthropic or open-source in 30 days if you had to? If not, you're vulnerable.

What To Do About It

Action Steps

Step 1: Audit Your Vendors. For each tool: vendor name, current use, military contracts, exit cost, reputational risk. Map it all.

Step 2: Diversify. Tier 1 (Mission-Critical): Anthropic or open-source. Tier 2 (Moderate): Mix OpenAI + Anthropic. Tier 3 (Commodity): Use OpenAI as a convenience.

Step 3: Build Governance. A simple Vendor Policy Committee. Quarterly meetings. Review vendor news. Audit tool usage. That's it.

Step 4: Tell Your Donors. ‘We've audited our vendors against our mission. We prioritise safety and transparency in our tool choices.’ They'll respect you more for thinking about this than for pretending it doesn't matter.

The Closing Question

The question isn't ‘Which tool is smartest?’

The question is: ‘Which tool lets me live according to my values?’

And if you can't answer that quickly, you're not being intentional about something that matters.

The market will shift if organisations demand it. Anthropic exists because customers prioritised safety over speed.

Your vendor choices are a statement of who you are. Choose wisely.

Ethics Briefing: How Your Tool Choices Reflect Your Values