# HiddenLayer joins DOE's $60M Prometheus project to secure nuclear AI > Source: > Published: 2026-08-18 14:23:08+00:00 [Chris Sestito](https://www.hiddenlayer.com/about-us?ref=runtimewire) is taking HiddenLayer's original thesis - that machine-learning systems require their own security controls - into the U.S. nuclear energy program. The Austin-based AI security company HiddenLayer said in [an August 18 announcement](https://www.prnewswire.com/news-releases/hiddenlayer-selected-to-support-does-60-million-prometheus-initiative-under-the-genesis-mission-302853514.html?ref=runtimewire) that it will contribute to Prometheus, a Department of Energy initiative led by Idaho National Laboratory. The project plans to use AI across reactor design, licensing, manufacturing, construction and operations, along with nuclear fuel fabrication and the management of legacy documents. Sestito founded HiddenLayer in 2022 with chief scientist Tanner Burns and CIO Jim Ballard after the three worked together in cybersecurity. HiddenLayer says the founders started the company after encountering a real-world adversarial-AI attack while working in cybersecurity, an experience that convinced them AI would become a distinct enterprise attack surface. That origin story gives HiddenLayer a logical role in Prometheus. Nuclear energy combines the two conditions that make AI security difficult: systems that can be manipulated in unfamiliar ways and infrastructure where an ordinary software failure can carry unusually high consequences. "AI has enormous potential to accelerate scientific discovery and transform critical industries," Sestito said in the announcement. He argued that deployment must be matched by a comparable commitment to security. ### A security seat, not a $60M check The [$60 million Phase II selection](https://inl.gov/news-release/genesis-mission-funds-ai-innovation-to-speed-up-safe-affordable-nuclear-energy/?ref=runtimewire) belongs to Prometheus as a whole and remains subject to congressional appropriations. HiddenLayer's release describes the company as a contributor to the collaboration; it does not characterize the entire amount as a contract or award to HiddenLayer. Idaho National Laboratory announced the selection on July 22, nearly four weeks before HiddenLayer identified its participation. INL described Prometheus as a 32-partner effort involving Oak Ridge, Argonne and Sandia national laboratories, academic institutions and more than 20 industry partners. Named participants include nuclear developers X-energy, TerraPower and Oklo. INL said the group had assembled more than $200 million in industry cost share and $30 million in industry capital, figures separate from the prospective federal funding. The project is the first Phase II award announced under the [Genesis Mission](https://www.energy.gov/undersecretaryforscience/genesis-mission/genesis-mission?ref=runtimewire), DOE's attempt to connect national laboratory computing, scientific facilities, AI systems and specialized datasets. DOE said in [July](https://www.energy.gov/undersecretaryforscience/articles/us-department-energy-announces-more-800-million-partner?ref=runtimewire) that the wider mission had secured more than $800 million in partner commitments, including compute resources, cloud infrastructure, model access, research support and direct funding. Prometheus has a narrower operating target: reduce the time and cost required to develop nuclear energy while preserving human involvement, technical review and established safety requirements. Those goals give HiddenLayer a government-led use case in which model behavior, data provenance and runtime activity require close scrutiny. HiddenLayer benefits from access to a government-led use case that is harder to dismiss as an AI security demonstration. Prometheus places AI inside regulated scientific and engineering workflows, giving Sestito's team a chance to apply its products where model behavior, data provenance and runtime activity require close scrutiny. ### From a Cylance attack to nuclear infrastructure Sestito previously led threat research at Cylance and later held engineering leadership roles at Qualys. Burns and Ballard also worked across Cylance, BlackBerry and Qualys, with backgrounds in malware research, threat classification and machine-learning security. The founders' first product focused on detecting and responding to attacks against machine-learning models without requiring access to the underlying training data or algorithms. When HiddenLayer [emerged from stealth in July 2022](https://www.hiddenlayer.com/news/hiddenlayer-launches-the-first-security-solution-to-protect-ai-powered-products?ref=runtimewire), Sestito said the team had founded HiddenLayer after helping respond to an attack directed through a machine-learning product. HiddenLayer raised a $6 million seed round around that launch. In September 2023, it [raised a $50 million Series A](https://www.hiddenlayer.com/news/hiddenlayer-raises-50m-in-series-a-funding-to-safeguard-ai?ref=runtimewire) co-led by M12, Microsoft's venture fund, and Moore Strategic Ventures. Booz Allen Ventures, IBM Ventures, Capital One Ventures and Ten Eleven Ventures also participated. HiddenLayer has publicly announced about $56 million in total funding; it has not attached a valuation to those rounds. HiddenLayer has since expanded beyond its original model-monitoring product. The current [AI security platform](https://hiddenlayer.com/?ref=runtimewire) covers discovery, supply-chain scanning, attack simulation and runtime protection for predictive, generative and agentic systems. Its [runtime documentation](https://docs.hiddenlayer.ai/docs/products/runtime/overview?ref=runtimewire) says the software monitors model inputs and outputs and can detect, redact or block malicious content. HiddenLayer describes newer agentic controls as reconstructing multi-turn sessions and tool calls so defenders can inspect an agent's activity as a connected sequence. That breadth matters because AI security has become a consolidation target for larger cybersecurity vendors. Palo Alto Networks completed its acquisition of Protect AI in July 2025 and later reported $634.5 million in purchase consideration. Protect AI's model scanning, red teaming and runtime capabilities now sit inside Palo Alto's Prisma AIRS offering. [Cisco's investor materials](https://investor.cisco.com/files/doc_presentation/2026/02/Cisco-Investor-Relations-Deck-2-19-26.pdf?ref=runtimewire) say the company acquired Robust Intelligence and incorporated its validation and runtime technology into AI Defense. HiddenLayer remains positioned as an independent specialist against those larger distribution channels. Prometheus gives Sestito another route: become embedded in the institutions defining how sensitive AI systems are tested and operated, rather than competing only through conventional enterprise security sales. ### The work starts where the announcement ends Prometheus spans several very different technical settings. Scanning historical nuclear documents presents different risks from protecting an AI system involved in reactor design, fuel fabrication or operational decision support. HiddenLayer will have to translate a broad platform pitch into controls that fit each workflow without obstructing the human review Prometheus has promised. The collaboration also raises the standard for evidence. Detecting a prompt injection in an office chatbot is a familiar security demonstration. Showing that the same defensive approach can identify relevant attacks, preserve auditability and operate within nuclear engineering constraints would carry more weight with government and critical-infrastructure buyers. Sestito started HiddenLayer after encountering an adversarial-AI attack while working in cybersecurity. Four years later, Prometheus gives the founders a place to test that thesis inside one of the country's most tightly controlled technical domains.