AI systems are increasingly integrated into cloud infrastructure, data analysis tools, and decision-support technologies used by governments worldwide. And recent conflict in the Middle East highlights the rapidly evolving role of artificial intelligence in military and national security settings.

For investors, the issue isn’t just whether technology companies work with government customers. Governments have long depended on private-sector tech providers. The more crucial question is how companies manage the use of their technologies in high-risk situations.

A common narrative suggests that once technologies are sold to governments, companies lose control over how they are used. In reality, technology providers retain significant leverage. Through contractual terms, technical architecture, and monitoring systems, companies can establish safeguards governing how their technologies are deployed.

The key question for investors is whether companies are exercising that responsibility.

Governance of High-Risk Technology Use

Responsible governance of AI in sensitive contexts requires companies to implement clear safeguards designed to prevent misuse and mitigate harm.

These safeguards generally include:

• Contractual restrictions on high-risk applications
• Robust customer due diligence processes
• Internal review procedures for sensitive government contracts
• Ongoing monitoring of product use after deployment
• Stakeholder engagement and grievance mechanisms
• The ability to suspend or terminate services when terms are violated

Importantly, due diligence in these contexts cannot be treated as a one-time exercise. High-risk deployments require continuous oversight and iterative monitoring, particularly when technologies may be used in conflict zones or national security operations.

These expectations are grounded in widely recognized international standards, including the UN Guiding Principles on Business and Human Rights and the OECD Guidelines for Multinational Enterprises. These frameworks establish that companies have a responsibility to identify, prevent, and address adverse human rights impacts linked to their operations and business relationships.

Contractual Safeguards in Practice

Recent developments involving major AI companies demonstrate how these safeguards can work in practice. Both Anthropic and OpenAI have included restrictions in their agreements regarding the use of their AI systems in military settings.

Anthropic sets boundaries on how the U.S. Department of Defense can deploy its Claude models. Recent tensions between Anthropic and U.S. government stakeholders have highlighted the practical challenges of enforcing these safeguards in defense-related contexts. Anthropic’s limits nevertheless reflect concerns commonly held by policymakers and military experts that current AI systems should not be used in fully autonomous lethal weapons. A retired three-star general who led the Pentagon’s Project Maven program has similarly warned that “no LLM, anywhere, in its current form, should be considered for use in a fully lethal autonomous weapon system.”9

OpenAI also includes operational safeguards, but it has taken a somewhat different, increasingly formalized approach as it has deepened its engagement with the U.S. government and defense-related stakeholders. The company emphasizes that its systems will primarily be deployed in managed cloud environments rather than distributed to local devices. This model reflects a broader industry shift toward enabling government use while retaining technical and contractual mechanisms for oversight and control.

OpenAI has also stated that security-cleared personnel should stay involved in deployment, enabling safety and alignment researchers to oversee and enhance systems over time. The company further notes that its agreements give it the right to end services if restrictions are violated, providing a means to enforce rules in the event of misuse.

OpenAI’s publicly disclosed contract provisions restrict certain applications, including:

• Fully autonomous lethal weapons
• Unconstrained domestic surveillance of U.S. persons
• Domestic law enforcement uses outside applicable legal frameworks

These provisions reference existing legal authorities, including the Fourth Amendment, the Foreign Intelligence Surveillance Act, Executive Order 12333, and the Posse Comitatus Act. While these legal standards already exist, incorporating them into contractual agreements adds extra enforcement tools and remedies.

Industry Precedent

The types of safeguards described above are not entirely new. Major cloud providers have long had Acceptable Use Policies that ban certain high-risk applications. For instance, the Google Cloud Platform restricts uses that could reasonably be expected to result in death, personal injury, or serious harm.

Similar provisions exist throughout the cloud industry, showing that technology companies have traditionally maintained some control over how their infrastructure is used. However, the rapid growth of AI capabilities has greatly increased the importance of these governance decisions.

AI Governance in Conflict Settings

Technology companies operating in conflict-affected or high-risk regions must also conduct heightened human rights due diligence.

The Global Network Initiative (GNI) has emphasized that companies should comply with international humanitarian and human rights law when operating in armed conflict zones. Companies should implement proactive processes to identify, prevent, and reduce potential harms throughout the lifecycle of their technologies. This includes evaluating risks before conflicts escalate, monitoring technology use during active conflicts, and providing support for remediation when harms happen. Traditional compliance processes are often inadequate in these settings. Companies need to update governance frameworks to handle quickly changing geopolitical risks.

Responding to Allegations

Recent developments involving Microsoft show how companies might react when concerns about technology use emerge. In 2025, allegations emerged that Microsoft’s Azure cloud infrastructure was being used to store surveillance-related data linked to military operations. Microsoft carried out both internal and independent reviews. After these reviews, the company found that some uses violated its terms of service and terminated services for a specific Israeli military unit.

Microsoft has also announced new governance measures, including expanded pre-contract review processes and reporting channels for employees raising human rights concerns. These actions highlight the importance of ongoing monitoring and the ability to take corrective steps when misuse is detected.

What Investors Should Focus On

As AI increasingly becomes part of defense, surveillance, and law enforcement systems, investors should demand greater transparency into how technology companies manage these risks.

Key questions include:

• What technologies are being provided to military/high-risk government users?
• What safeguards govern those deployments?
• How is technology use monitored after deployment?
• How are allegations of misuse responded to and investigated?
• Under what conditions would companies suspend or terminate services?

Investors should also expect companies to align governance practices with international standards, including the UN Guiding Principles on Business and Human Rights and relevant international humanitarian law.

Artificial intelligence is advancing rapidly, and the geopolitical environment in which it is used is becoming increasingly complex. Responsible management thus requires transparency, accountability, and ongoing oversight. For investors focused on long-term value and responsible innovation, understanding how companies handle these risks is becoming an increasingly important part of engaging with the technology sector.

[1] Shanahan, Jack, quoted in coverage of the Anthropic–Pentagon Dispute Sparks Debate on Limits of Battlefield AI | February 2026, Bloomberg