SP
BravenNow
The Pentagon is planning for AI companies to train on classified data, defense official says
| USA | technology | ✓ Verified - technologyreview.com

The Pentagon is planning for AI companies to train on classified data, defense official says

#Pentagon #classified data #AI training #generative AI #military intelligence #security clearances #warfighting

📌 Key Takeaways

  • The Pentagon plans secure environments for AI companies to train models on classified military data.
  • Training on classified data aims to improve AI accuracy for tasks like target analysis and battlefield assessments.
  • AI firms may access classified data under strict security protocols, though the DoD retains data ownership.
  • The initiative aligns with the Pentagon's push to become an 'AI-first' warfighting force amid escalating conflicts.

📖 Full Retelling

The Pentagon is discussing plans to set up secure environments for generative AI companies to train military-specific versions of their models on classified data, MIT Technology Review has learned. AI models like Anthropic’s Claude are already used to answer questions in classified settings, including for analyzing targets in Iran. But allowing models to train on and learn from classified data would be a new development that presents unique security risks. It would mean sensitive intelligence like surveillance reports or battlefield assessments become embedded into the models themselves, and bring AI firms into closer contact with classified data than before. Training versions of AI models on classified data is expected to make them more accurate and effective in certain tasks, according to a US defense official who spoke on background with MIT Technology Review . The news comes as demand for more powerful models is high: the Pentagon has reached agreements with OpenAI and Elon Musk’s xAI to operate their models in classified settings, and is implementing a new agenda to become an “an ‘AI-first’ warfighting force” as the conflict with Iran escalates. (The Pentagon did not comment on its AI training plans as of publication time.) Training would be done in a secure data center that’s accredited to host classified government projects, and where a copy of an AI model is paired with classified data, according to two people familiar with how such operations work. Though the Department of Defense would remain the owner of the data, personnel from AI companies with appropriate security clearances might in rare cases access the data, the official said. Before allowing this new training, though, the official said the Pentagon intends to first evaluate how accurate and effective models are when trained on non-classified data, like commercially available satellite imagery. The milita

🏷️ Themes

Military AI, Data Security

📚 Related People & Topics

The Pentagon

The Pentagon

Headquarters of the U.S. Department of Defense

The Pentagon is the headquarters building of the United States Department of Defense, in Arlington County, Virginia, across the Potomac River from Washington, D.C. The building was constructed on an accelerated schedule during World War II. As a symbol of the U.S. military, the phrase The Pentagon i...

View Profile → Wikipedia ↗

Entity Intersection Graph

Connections for The Pentagon:

🌐 Silicon Valley 1 shared
🌐 Artificial intelligence 1 shared
🌐 Pentagon 1 shared
View full profile

Mentioned Entities

The Pentagon

The Pentagon

Headquarters of the U.S. Department of Defense

Deep Analysis

Why It Matters

This development matters because it represents a significant shift in military AI capabilities, potentially giving the U.S. military more accurate and effective AI tools for battlefield decision-making and intelligence analysis. It affects national security by embedding classified intelligence directly into AI models, which could create new vulnerabilities if models are compromised. The move also brings private AI companies into closer collaboration with military operations, raising ethical questions about the militarization of commercial AI technology. This could accelerate an AI arms race as other nations respond with similar initiatives.

Context & Background

  • The Pentagon has already been using AI models like Anthropic's Claude in classified settings for tasks such as analyzing targets in Iran
  • The Department of Defense has existing agreements with OpenAI and Elon Musk's xAI to operate their models in classified environments
  • The U.S. military has been implementing an 'AI-first' warfighting agenda as tensions with Iran escalate
  • Current AI models in military use are typically trained on non-classified data and then deployed in secure environments
  • There's growing global competition in military AI development with China and Russia pursuing similar capabilities

What Happens Next

The Pentagon will likely conduct pilot programs testing AI models trained on non-classified data first, such as commercially available satellite imagery, to evaluate effectiveness before proceeding with classified data training. Expect increased scrutiny from Congress and defense oversight committees about security protocols for protecting classified data within AI models. International reactions may include similar initiatives from allied nations and countermeasures from adversaries like China and Russia. Look for potential protests or ethical debates about the militarization of commercial AI technology.

Frequently Asked Questions

What security risks does training AI on classified data create?

Training AI on classified data creates risks of sensitive intelligence becoming embedded in the models themselves, potentially making them targets for adversaries seeking to extract classified information. There's also the risk of data leakage through model outputs or if company personnel with security clearances mishandle the information. The models could become single points of failure containing vast amounts of sensitive military intelligence.

How will AI companies access classified military data?

AI companies will operate within secure data centers accredited for classified government projects, where their models will be paired with classified data. Company personnel would need appropriate security clearances and would only access data in rare cases under strict protocols. The Department of Defense would maintain ownership and control over all classified data throughout the process.

Why does the Pentagon want AI trained on classified data?

The Pentagon believes training AI models on classified data will make them more accurate and effective for military-specific tasks like battlefield assessment and intelligence analysis. This could provide tactical advantages in conflicts like the escalating tensions with Iran. It aligns with the military's goal of becoming an 'AI-first' warfighting force with more sophisticated decision-support systems.

Which AI companies are involved with the Pentagon?

The Pentagon already has agreements with OpenAI and Elon Musk's xAI to operate their models in classified settings. While not confirmed for the new training initiative, companies like Anthropic (maker of Claude) are already used in classified military operations. The program would likely involve multiple leading AI firms with appropriate security capabilities.

What safeguards will protect the classified data?

Training will occur in secure, accredited data centers designed for classified government projects. The Department of Defense will retain ownership of all data, and AI company access will be strictly limited to cleared personnel in rare circumstances. The Pentagon plans to first test models on non-classified data to evaluate effectiveness before proceeding with classified training.

How might this affect international AI competition?

This initiative could accelerate global military AI development as other nations respond with similar programs, potentially creating an AI arms race. Allies may seek similar collaborations with their domestic AI companies, while adversaries like China and Russia will likely intensify their own military AI efforts. It could also influence international norms around military applications of commercial AI technology.

}
Original Source
The Pentagon is discussing plans to set up secure environments for generative AI companies to train military-specific versions of their models on classified data, MIT Technology Review has learned. AI models like Anthropic’s Claude are already used to answer questions in classified settings, including for analyzing targets in Iran. But allowing models to train on and learn from classified data would be a new development that presents unique security risks. It would mean sensitive intelligence like surveillance reports or battlefield assessments become embedded into the models themselves, and bring AI firms into closer contact with classified data than before. Training versions of AI models on classified data is expected to make them more accurate and effective in certain tasks, according to a US defense official who spoke on background with MIT Technology Review . The news comes as demand for more powerful models is high: the Pentagon has reached agreements with OpenAI and Elon Musk’s xAI to operate their models in classified settings, and is implementing a new agenda to become an “an ‘AI-first’ warfighting force” as the conflict with Iran escalates. (The Pentagon did not comment on its AI training plans as of publication time.) Training would be done in a secure data center that’s accredited to host classified government projects, and where a copy of an AI model is paired with classified data, according to two people familiar with how such operations work. Though the Department of Defense would remain the owner of the data, personnel from AI companies with appropriate security clearances might in rare cases access the data, the official said. Before allowing this new training, though, the official said the Pentagon intends to first evaluate how accurate and effective models are when trained on non-classified data, like commercially available satellite imagery. The milita
Read full article at source

Source

technologyreview.com

More from USA

News from Other Countries

🇬🇧 United Kingdom

🇺🇦 Ukraine