Axonius, Fortanix target AI security gaps with autonomous workers and confidential containers

Axonius said AI Workforce lets security teams hand ongoing remediation work to autonomous, AI-powered virtual workers. The company argues scheduled jobs catch only conditions someone defined in advance, while asset data is split across dozens of tools and exceptions are left for people to sort out. The virtual workers come in more than 50 types, covering exposure triage, shadow AI and readiness for post-quantum cryptography, and each tracks a metric tied to its outcome. When an asset drifts from its intended state, the responsible worker carries out the fix and sends results into a Daily Digest ranked by urgency.

A mode setting controls autonomy. In Off mode, a worker waits for approval on everything; Shadow mode previews what it would do; Live mode lets it act automatically within guardrails. Axonius said outbound messages and changes to connected systems are treated as sensitive and held for approval. Each customer runs a private large language model instance in a dedicated environment, and an audit trail marks every action as agent-driven or human-driven. AI Workforce, AI Companion and MCP Server run on the same asset model used by Asset Cloud, which reconciles and deduplicates data across more than 80 asset and exposure entity types and draws on more than 1,400 bidirectional integrations. AI Companion accepts plain-language questions and can open a ticket or tag devices in the same conversation; MCP Server lets outside AI tools and agents reach the same asset data over Model Context Protocol.

Moshe Ben Simon, Axonius's chief product officer, said every vendor is shipping AI agents and "the question nobody is answering is what those AI agents are acting on." An agent acting on one vendor's incomplete view of an asset will be wrong, he said. Theresa Lanowitz, principal analyst for cybersecurity at Omdia, said the launch's standout is how directly it ties autonomous remediation to a verified view of the environment. AI Workforce is generally available now, and security engineering teams can see AI credit spending by users and virtual workers.

Fortanix's update targets a different exposure: during inference, model weights and input data must be decrypted in processor memory, where privileged administrators and cloud operators with infrastructure access may reach them. Fortanix said it extended Confidential AI to support Confidential Containers, which reached incubating status at the CNCF in July, so AI workloads on Kubernetes can run inside hardware-based trusted execution environments. Open-source plugins verify a container before it approaches an encryption key. The check runs through Fortanix Confidential Computing Manager, which attests CPU and GPU together so an inference job spanning both is covered, and Data Security Manager releases keys only if the result meets the customer's security policy.

Fortanix published survey figures last week on how AI leaders view the risk. Atomik Research polled 300 AI leaders at large U.S. companies on Fortanix's behalf; 93% said traditional encryption does not protect data while AI is processing it, and 41% said security concerns had led them to block a deployment completely. Co-founder and Chief Executive Anand Kashyap said moving AI into containers "doesn't change the threat," since models and agents still touch sensitive data while they run. Fortanix said deploying a container into a confidential computing environment takes one click on supported hardware and Kubernetes platforms, and key release runs on a cluster so losing one server does not block it. Customers with sovereignty requirements can run the platform on-premises or fully air-gapped, which Fortanix says can let an enterprise host a frontier lab's proprietary model next to its own data without putting the lab's intellectual property at risk.

Anuj Jaiswal, Fortanix's chief product officer, said "Attestation is a protocol. Trust is an operation," and that having attestation and key management on one platform spares enterprises from building the trust layer themselves. AMD, Intel and CrowdStrike are among more than a dozen companies Fortanix lists as supporting the work. John Fanelli, Nvidia's vice president of enterprise software, said Fortanix is building on open standards for attestation and secure key release "at production scale." Fortanix Confidential AI for Confidential Containers is available now; the company plans to demonstrate it on Nvidia GPUs with confidential computing enabled at Nvidia GTC Berlin from Oct. 20 to 22.