Today, Eclypsium released new research demonstrating the importance of addressing security risks in the infrastructure used for training and running AI models and agents, as well as the complex supply chains behind AI infrastructure components.

If left unaddressed, infrastructure-level vulnerabilities can allow threat actors to steal and manipulate AI models during live inference or impact the availability of the AI infrastructure itself.

This research underscores the importance of building specialized security capabilities and solutions focused on protecting AI infrastructure and its supply chains.

Eclypsium is working with NVIDIA to advance infrastructure assurance through integration with the Open Agent Safety Platform.

Why It Matters

Global investment in AI infrastructure is projected to exceed $1 trillion annually, with AI expected to contribute $15.7 trillion to global GDP. Yet today’s assurance practices focus almost exclusively on model safety and governance, not the underlying hardware and firmware that determines whether a model executes faithfully.

Eclypsium’s findings highlight the importance of infrastructure protections: AI models may be compromised while the model checks show no signs of tampering.

Eclypsium’s Infrastructure Assurance solution provides independent verification and continuous monitoring of AI infrastructure.

Key Findings

  • An attack chain targeting a privileged component of AI compute infrastructure, enabling compromise of a model including weights

  • Ability to alter model output during live inference without modifying the stored model or inference service

  • No visible indicators of compromise

  • Controlled demonstration showing manipulated model output on queries

Working with NVIDIA

Eclypsium is also working with NVIDIA to advance infrastructure security through research and collaboration on continuous, independent verification of infrastructure. Both NVIDIA and Eclypsium support a shared objective: building and maintaining a stronger chain of trust, from verified hardware and firmware to independently enforced AI agent execution.

Assets Available

  • Blog by NVIDIA and Eclypsium about the importance of AI Infrastructure Assurance and their collaboration

  • New technical research detailing risks and threats to AI infrastructure and their impact on AI models and agents, including published vulnerabilities and available mitigations

Executive & Expert Quotes

Yuriy Bulygin, CEO and co-founder of Eclypsium:

“AI security starts with trusting the infrastructure that it runs on. Our research demonstrates how hardware-level compromises can undermine AI execution without being detected by the stored-model integrity checks used in the demonstration. Through Eclypsium’s independent verification and continuous monitoring, we’re helping organizations establish and maintain trust in the infrastructure powering AI.”

Gilad Shainer, SVP of Networking at NVIDIA:

“NVIDIA builds security into its platforms to help customers protect their AI infrastructure. Working with researchers and cybersecurity companies, like Eclypsium, we’re continuously strengthening protections across the AI ecosystem.”

Jon Rolf, cybersecurity consultant at Rolf Cyber LLC and retired Director of the US Common Criteria evaluation scheme at NSA:

“Some of the most consequential and complex decision systems ever are being developed now. The new architectures that the AI systems are running on due to complexity and speed benefit from independent verification. AI infrastructure should be treated as critical infrastructure with verifiable evidence that it’s in a known-good state. Securing critical infrastructure is a team sport and this is a successful example of how to do it: researchers demonstrated an attack scenario and worked with the manufacturer to help clarify practical protections that the whole industry can use.”

Dr. Allan Friedman, former Senior Advisor and Strategist at CISA:

“This research reinforces the need for full-stack transparency and monitoring across AI infrastructure. Organizations can’t secure what they can’t see, and the same transparency standards applied to software must extend to firmware and hardware. A Bill of Materials is just the beginning. AI infrastructure needs to turn this transparency into continuous evidence so that we can discover and address security problems.”

Media gallery

About The Author
Business Wire