AI hallucinations are emerging as a critical barrier to enterprise cybersecurity adoption. Security leaders increasingly question whether generative AI tools can be trusted to support high-stakes operational decisions, particularly in cloud and hybrid environments where fragmented data, opaque reasoning, and alert fatigue already dominate workflows.
Against that backdrop, Uptycs has launched Juno AI Analyst, an agentic investigation tool designed to address what the company describes as the primary constraint on AI adoption in Cloud-Native Application Protection Platforms — trust. Uptycs says Juno has already been deployed by automotive manufacturers, banks, and large enterprises, replacing earlier-generation AI tools focused on compliance reporting rather than operational verification.
Unlike common security “AI copilots” that summarise alerts from multiple products, Juno is positioned as an autonomous investigator. The platform executes deterministic SQL queries against Uptycs’ unified, multi-cloud data model, enabling it to surface raw telemetry as evidence for each conclusion it produces. The company refers to this as a “glass box” approach, intended to replace probabilistic outputs with transparent reasoning that can be inspected and verified by security teams.
“Most of the time, uncertainty rules in the world of cybersecurity,” said Ganesh Pai, CEO of Uptycs. “Juno’s evidence-based approach uses AI to replace opaque ‘black box’ answers with transparent, verifiable reasoning grounded in real telemetry, so security teams can trust what they’re seeing and act with confidence.”
The release of Juno comes as security operations teams face growing pressure to consolidate tooling and improve response times. Industry analysts have repeatedly highlighted the limitations of federated security architectures, where cloud, endpoint, identity, and workload data sit in separate databases stitched together through APIs. While these models enable broad coverage, they often prevent AI systems from reliably correlating activity across environments.
Uptycs argues that this architectural fragmentation is a root cause of AI hallucinations in security tooling. By normalising telemetry into a single schema — comprising thousands of tables and hundreds of thousands of data fields — Juno is able to query precise slices of data rather than ingesting large volumes of loosely connected information. The company says this reduces noise, limits false inference, and allows findings to be traced back to original sources such as CVE records and vendor documentation.
Enterprise partners appear to be responding to that proposition. Srinivas Tummalapenta, CTO of IBM CyberSecurity Services and an IBM Distinguished Engineer, said the platform reflects a broader shift in how security professionals want to interact with AI. “Juno addresses the cybersecurity professional’s aspiration to leverage a conversational interface, interact with their own data, and harvest insights, explanations, and recommendations,” he said. “This capability increases the engagement of cyber professionals and further accelerates AI adoption and automation implementation in cyber defense and response mechanisms.”
Channel partners are also positioning Juno as a response to growing buyer scepticism around AI-led security claims. Defy Security CRO Rich Douros said customers are increasingly asking how AI systems reach their conclusions, not simply what actions they recommend. “Its ability to provide clear, verifiable insights, rather than opaque alerts, is something that resonates strongly in the market,” he said.
As regulators and boards scrutinise the operational risks of autonomous systems, the emphasis on explainability is becoming central to enterprise technology procurement. In cybersecurity, where automated decisions can disrupt production systems or critical infrastructure, vendors that can demonstrate traceability and evidence may hold a structural advantage as AI moves deeper into security operations.
You must be logged in to post a comment.