Widespread AI-generated disinformation—even if aimed “only” at foreign populations—inevitably backfires. Internet borders are porous.

The U.S. adopting this strategy mirrors tactics used by adversarial regimes. As Heidy Khlaaf of the AI Now Institute noted, “offensive and defensive uses are really two sides of the same coin.”

The U.S. Military’s AI Propaganda Ambitions – How It Works, Why It Matters, and How to Defend Against It

by ChatGPT-4o

---

I. Overview and How It Would Work

A recent Pentagon wishlist, exposed by The Intercept, reveals that the U.S. Special Operations Command (SOCOM) is actively seeking to acquire AI-enabled propaganda systems to influence foreign populations and "suppress dissenting arguments." The request outlines a desire to deploy agentic AI systems and large language models (LLMs)—like OpenAI's ChatGPT or Google’s Gemini—for real-time narrative control, foreign influence campaigns, and disinformation at scale.

Here’s how the envisioned system would function:

1. Automated Scraping & Analysis
   AI agents would scrape online conversations, analyze trends, emotional tone, and political sentiment, then determine how best to intervene or shape narratives.

2. Autonomous Content Generation
   Using LLMs, these systems would generate persuasive, context-specific propaganda—tailored to culture, language, psychology, and even individual personalities—to manipulate discussions.

3. Simulation of Social Responses
   The Pentagon seeks to test these messages against agent-based replicas of target populations to simulate and optimize outcomes—essentially digital sandboxes to test manipulation strategies.

4. Targeting Critics
   Alarmingly, the tech would also track and profile those who oppose U.S. messaging, creating hyper-targeted counter-narratives or suppressive responses to silence dissenting voices.

5. Deepfake & Influence Expansion
   Offensive deepfake capabilities remain part of the toolbox, supporting psychological operations with fake videos or personas for enhanced deception and plausible deniability.

---

II. Positive or Negative Development?

This is a profoundly dangerous development—ethically, strategically, and socially.

Negative Aspects:

• Erosion of Trust and Democracy: Widespread AI-generated disinformation—even if aimed “only” at foreign populations—inevitably backfires. Internet borders are porous. Such tools may spill into U.S. discourse or be used by future administrations against domestic dissent.

• Authoritarian Precedent: The U.S. adopting this strategy mirrors tactics used by adversarial regimes. As Heidy Khlaaf of the AI Now Institute noted, “offensive and defensive uses are really two sides of the same coin.”

• Mission Creep: Previous operations (e.g., the anti-vax Sinovac campaign revealed by Reuters) show that supposed "foreign influence efforts" often end up targeting civilians, journalists, or allies.

• Technical Risks: These systems are unreliable. LLMs fabricate, misunderstand context, and reflect user bias—making propaganda based on them error-prone, manipulable, and potentially counterproductive.

The Pentagon’s Defense:

SOCOM claims adherence to the DoD’s Responsible AI framework and insists these tools won’t be used on American citizens. But historical precedent shows that intent and outcome often diverge, especially in the fog of cyberwar.

---

III. How Countries, Businesses, and Civilians Can Protect Themselves

In light of this disturbing revelation, prevention, resilience, and transparency are crucial. Here’s what different groups should do:

1. For Countries:

• Establish International Treaties or Norms: Push for a Geneva Convention-style framework for AI propaganda and information warfare, to prevent the proliferation of digital psyops across borders.

• Implement Stronger Attribution Protocols: Develop AI watermarking, cryptographic proofs, and traceability protocols to detect and expose synthetic propaganda content.

• Promote Media Literacy & Digital Hygiene: National education systems must teach critical thinking, fact-checking, and psychological resilience to digital influence.

2. For Businesses and Platforms:

• Develop Counter-Propaganda AI: Social media platforms and news aggregators should invest in AI tools that detect coordinated inauthentic behavior, deepfakes, and linguistic manipulation.

• Transparency Obligations: Platforms must label AI-generated content, disclose influence operations they detect, and publish regular transparency reports.

• Resist Co-option: Companies must resist government or military demands that conflict with human rights, democratic values, or global trust.

3. For Civilians and Civil Society:

• Verify Before Sharing: Be cautious of emotionally charged content, especially from unknown sources. Use fact-checking tools and avoid knee-jerk reposting.

• Support Watchdogs: NGOs like Freedom House, EFF, Access Now, and The Intercept play a crucial role in uncovering government overreach. Support them through engagement and funding.

• Advocate for Regulation: Citizens should demand accountability and AI transparency legislation, especially regarding how AI is used by military, intelligence, and law enforcement bodies.

---

IV. Conclusion

The SOCOM document is not merely a glimpse into speculative military tech—it is a blueprint for algorithmic influence at societal scale. If allowed to proceed unchecked, the deployment of agentic AI to suppress dissent could spell disaster not just for foreign societies, but for global trust in democratic information systems. AI may be the new battlefield, but truth is the first casualty.

To preserve democracy, digital sovereignty, and open discourse, governments and civil societies must move fast—not only to regulate this new form of warfare but to immunize against it.

“AI tends to make these campaigns stupider, not more effective,” said Emerson Brooking. Unfortunately, when the stakes are high enough, even stupid propaganda can kill. The world must not wait until it does.