- VAST Data is launching PolicyEngine and TuningEngine to help organizations deploy AI that's "secure, trusted and self-learning," the company said
- The technologies help make AI fit for business, said Futurum Group analyst Brad Shimmin
- VAST, an up-and-coming contender in data platforms and AI, kicks off its VAST Forward partner and customer conference this week
VAST FORWARD 2026, Salt Lake City — AI's history of errors is horrifying, or hilarious, if you have a dark sense of humor. A Microsoft travel site listed a Canadian food bank as a hot tourist destination and encouraged, "Consider going into it on an empty stomach!" A teacher gave his entire class an "incomplete" grade after AI falsely accused students of having used AI on their final assignments. Two U.S. lawyers were fined for submitting fake court citations from ChatGPT. And when Google launched its Bard AI chatbot, the demo prominently featured the AI getting answers wrong, sending Alphabet stock sliding downward.
To be fair, those examples are three years old — AI has gotten a lot better. But AI still requires extensive fact-checking. And people are much more concerned than excited about AI, according to a September Pew study.
VAST Data is looking to close the AI trust gap. At this week's VAST Forward conference for customers and partners, VAST is unveiling tools to help organizations deploy AI that's "secure, trusted and self-learning," the company said.
VAST's new PolicyEngine is a governor that provides guardrails for agentic activity, while TuningEngine manages model tuning. The two tools are designed to work in conjunction to help AI learn within guardrails.
"One of the primary blockers to broad adoption of AI within organizations is trust in the models — trust in what the models have been trained upon, trust in what the models are allowed to do with different data or different tools and then ultimately trust in agents that use these models," VAST Data co-founder Jeff Denworth said at a Zoom briefing for press and analysts last week. "The thinking machine needs data and needs to be trusted."
Closing the trust gap
VAST's new tools are designed to help close that gap. “With the introduction of PolicyEngine and TuningEngine, the VAST AI Operating System has become a thinking machine that customers can deploy wherever they compute – a machine that safeguards every interaction and learns from every outcome, bringing the power of AI within reach of every organization," Denworth said.
The VAST PolicyEngine provides controls on what information agents can access and how they communicate with other agents as well as tools and remote data products to control data spillage and leakage. The service is necessary as AI workflows and agents access organizational data and use it to produce more information in the form of generated responses, agent-to-agent communications, event logs and more.
The TuningEngine captures outcomes from agentic pipelines and uses curated feedback to enhance model performance over time. It uses popular methods such as LoRA fine tuning, supervised fine tuning and reinforcement learning to ingest data, process it and suggest new candidate models, which can then be evaluated and deployed to kick off a new learning loop to further improve models.
Both tools are slated for release by the end of the year.
A layer in the AI cake
Founded in 2016, VAST is an up-and-coming enterprise AI player. It provides an AI Operating System, a platform that consolidates data from multiple sources — files, objects, tables, vectors, streams and metadata — and uses that data to run AI agents at enterprise scale with secure runtime, governed access and full auditability.
The company sees its mission — its "north star" — as building a "thinking machine," said Renen Hallak, founder and CEO, at an in-person briefing for press and analysts Tuesday evening.
That's also the mission for OpenAI, Anthropic, Google Gemini and Elon Musk's X, and other big AI companies. How is VAST different? "We're not in any way going to do it on our own," Hallak said. He noted that Jensen Huang, president and CEO of Nvidia (a VAST partner and investor) has said that AI is a "five-layer cake:" Energy, chips, cloud services, models and applications. "It's a new stack. And you need all the parts of the new stack for this to work." VAST provides the software and cloud services layers for AI, and partners with Nvidia and other chip vendors on hardware. And several of the other companies in the stack build using VAST, Denworth said.
VAST has more than $4 billion cumulative bookings since inception, more than $500 million committed annualized run rate coming out of fiscal 2026, with annual-sales roughly tripling year-over-year and a self-sustaining business with significant free cash flow and positive operating income, Denworth said. "Never have you found a company growing at the pace we are that is not burning mountains of venture capital," Denworth said.
VAST works with major cloud players like Google Cloud, CoreWeave, Vultr as well as Nvidia, which is also one of their investors.
VAST is, essentially, a storage company, said Futurum Group analyst Brad Shimmin. They compete with a murder's row of platform providers, including Dell, Everpure (formerly known as Pure Storage — the company rebranded this week), Hewlett Packard Enterprise, NetApp, MinIO, Google, Amazon Web Services (AWS) and Microsoft Azure.
Putting data to 'practical work'
VAST's latest updates enhance AI's business fitness, Shimmin said. "They are focusing on things that tend to get lost in this gold rush that we're in right now for agentic system: Practical machine learning and the art of data science," he said. "This is not about who can build the best prompt or agentic harness, but instead who can take data — the data that defines a business — and put it to practical work in training and informing a large language model such that whoever comes in contact with that model experiences an accurate representation of the business."
The PolicyEngine will help enterprises build smaller, more performant, more transparent and less expensive language models specialized for their needs, Shimmin said. He noted that AWS is pursuing a similar goal with its Nova Forge technology, introduced in December.
The VAST technology can help enterprises meet demands for data sovereignty, by deploying AI locally, Shimmin said.
Industry partnerships for AI
Also this week, VAST introduced Polaris, a global control plane designed to provision, operate and orchestrate distributed AI infrastructure across public cloud, neocloud and on-premises data center environments, enabling enterprises to manage AI data and infrastructure wherever training and inference workloads run.
Additionally, VAST announced a partnership with TwelveLabs for intelligence to power video search, analytics and reasoning at massive scale. TwelveLabs allows AI systems to describe what's happening in a video in natural language and enables users to search video archives, also in plain language. VAST described the capability using this illustration:
TwelveLabs has primarily been cloud-based; VAST helps the company expand into markets where video content needs to be kept near the source or on premises.
Additionally, VAST unveiled a partnership with Nvidia to introduce an end-to-end fully accelerated AI data stack, using Nvidia libraries to accelerate both compute and data services for RAG, vector search, real-time SQL and agentic applications, running on GPU accelerated servers.
The Nvidia partnership follows an announcement Tuesday where Nvidia partnered with Red Hat to launch a turnkey AI factory that combines Red Hat software with Nvidia hardware and models.
VAST also launched a partnership with CrowdStrike for improved security across the VAST AI Operating System, embedding CrowdStrike's enterprise-grade threat detection and automated response capabilities.
And VAST extended its Cosmos Community, where AI practitioners collaborate and share knowledge. The new Cosmos partner program provides resources to help partners build, validate, deliver and scale customer solutions on the VAST AI Operating System. These partners include software, hardware, solution and cloud providers, developers and value-added resellers.