Cranium AI, an AI security and governance company based in Short Hills, New Jersey, announced on May 21, 2026, that it had acquired Aiceberg, a New York startup that builds security and risk tools for AI applications and agents. Terms were not disclosed.
Aiceberg chief executive Alex Schlager joined Cranium as chief technology officer. He is responsible for Cranium's technical roadmap and for merging the two companies' technology. All of Aiceberg's employees moved to Cranium, bringing staff in AI security, data science, engineering and sales.
Cranium said the combined company would offer one platform covering the full life of an AI system, from development to the deployment of autonomous agents.
What Aiceberg built
Aiceberg launched its platform in March 2025 alongside its seed round. It works as a firewall and gateway between users and AI models: it inspects prompts going in and responses coming out, including text, speech, images and source code, and enforces security and company policies in real time.
The product looks for attacks such as prompt injection, prompt leaking and jailbreaking, blocks off-topic or unsuitable requests, and redacts personal and sensitive information. It also includes controls on the actions AI agents can take. Aiceberg stressed that it uses its own non-generative models to detect risk, rather than relying on large language models to police other language models.
In the acquisition announcement, Cranium described Aiceberg's main contribution as technology that maps risk across AI agents.
Schlager said at the seed launch that Aiceberg works independently of the AI applications it protects, analysing only the content of inputs and outputs. The company said its explainable, non-generative models made each decision auditable from start to finish.
“I am thrilled to join as CTO and lead our combined talented teams as we build the essential trust layer for the modern enterprise.”
Why Cranium is buying
Cranium sells software that helps companies discover, monitor and secure the AI models and systems they use. It said the deal addresses three needs as companies move from AI experiments to agent-based workflows: protecting language models and generative applications against attacks, tools to monitor and control autonomous agents, and automated mapping of AI systems to regulatory standards.
Cranium chief executive and co-founder Jonathan Dambrot said bringing in Aiceberg's team and technology would speed up Cranium's work on a trust layer for enterprise AI. The companies said the combination would create the largest independent platform focused on securing and governing agentic AI systems.
Cranium, headquartered in the New York metropolitan area, describes itself as an end-to-end platform for securing and governing AI across the full model lifecycle. It has also announced an integration with Weights & Biases, now part of CoreWeave, to make AI safety and security testing a standard step in model development, and a partnership with cyber-resilience adviser ISTARI.
The founders and the funding
Aiceberg announced a $10 million seed round on March 6, 2025, led by SYN Ventures and Sprout & Oak. SYN Ventures invests in security companies, and Sprout & Oak is an early-stage firm focused on AI startups. The company planned to use the money for real-time risk detection and to expand its security and compliance features. No later funding was reported, and Aiceberg's headcount was not disclosed.
SYN Ventures had also backed Cranium. Cranium was spun out of KPMG's startup incubator, KPMG Studio, in 2023 with seed money from KPMG and SYN Ventures, and in October 2023 raised a $25 million round led by Telstra Ventures with participation from KPMG and SYN Ventures, ROI-NJ reported.
Dambrot said at the time of the Series A that companies should prioritise responsible AI early to avoid compliance problems later. Cranium said the 2023 funding would go to product development, sales and marketing and research into the security of AI systems.
A busy market for AI security
Aiceberg's argument, set out when it launched, was that many organisations had a false sense of security because the tools they used to check language models were themselves built on language models.
By folding Aiceberg into its platform, Cranium adds runtime protection for AI applications and agents to its existing governance and compliance tools.
