The news
RSA and Radware (RDWR) launched separate AI agent identity products on September 29, 2026. The launches came a day after SiliconANGLE reported an Omdia survey showing security leaders are uncertain where agent threats will come from and split on which platform should handle agent security.
RSA said its new Agent ID platform is built for government, financial services, critical infrastructure and other highly regulated sectors, and announced it at The AI Conference in San Francisco. RSA CEO Greg Nelson said that in these sectors, getting agentic security wrong "is not inconvenient, it is catastrophic." According to the company, a Discover module finds known and unknown agents and Model Context Protocol (MCP) servers, the connectors agents use to reach tools and data, across identity, cloud, endpoint and gateway sources.
A Secure module enforces policy at a gateway and requires a named, authenticated operator to approve high-risk actions out of band, using a phishing-resistant credential, while a Govern module runs continuous certification and risk-based access reviews. RSA said gateways can run in the cloud, in a hybrid setup or on premises, that tenant data stays in the US or EU, and that enforcement records stream to customers' security information and event management (SIEM) systems. A fully air-gapped, self-managed version is planned for 2027. The release did not give pricing.
Radware's Agent Trust Management, announced the same day and available now as part of its Cloud Application Protection services, targets a different problem: outside AI agents that visit a company's websites and applications. Radware said the product captures each agent's declared intent, uses behavioral analysis and anomaly detection to check whether its activity stays consistent with that intent, and gives companies fine-grained control, for example letting an agent view pages but blocking it from signing in to an account or completing a purchase.
Radware cited its own H1 Global Threat Analysis Report as finding that 77% of organizations are deploying or implementing AI agents, while only 17.2% report full visibility into agent operations. Radware Chief Technology Officer David Aviv said the product is meant to help companies decide which agents to trust and limit the actions those agents are allowed to take.
The Omdia findings came from Todd Thiemann, Omdia's principal analyst for identity and access management and data security, in an interview with theCUBE Research at Okta's Oktane 2026 conference, SiliconANGLE reported on September 28. In Omdia's survey of about 400 security leaders, the top barrier to adopting identity security for AI agents was confusion and uncertainty about where attacks will come from; SiliconANGLE's report did not describe when or how the survey was conducted. On which platform should handle agent security, Thiemann said the most common answer was a data security platform, with an identity platform next.
The numbers
- Security leaders in Omdia's agent identity survey
- About 400
- Organizations deploying or implementing AI agents (Radware report)
- 77%
- Organizations reporting full visibility into agents (Radware report)
- 17.2%
- Planned air-gapped version of RSA Agent ID
- 2027
Why CEOs should care
For CISOs, the first step is to separate two problems that vendors both describe as agent identity. One is governing the agents your company runs, which is where RSA's product sits: which agents exist, what they can reach and who approves risky actions. The other is managing outside agents that use your customer-facing applications, which is Radware's focus. Ownership and budget should follow that split, since the first usually belongs to identity teams and the second to application security.
For buyers choosing between a platform suite and a point tool, Omdia's survey points to no settled answer, so the criteria matter more than the category. Ask whether a product finds agents and MCP servers nobody sanctioned, whether it enforces policy at runtime or only reports, and whether it produces evidence an auditor would accept, such as a named operator approving each high-risk action. Ask where data is stored, whether it can run on premises, and how it works with the identity provider you already use.
For CFOs and boards, consolidation can cut vendor count but may lock in a stack before the market settles. Boards in regulated industries should ask management which agents can take consequential actions today, who approves those actions, and whether the record would satisfy a regulator.
The bigger picture
The field is filling fast. SiliconANGLE noted that Okta (OKTA) recently added an AI agent runtime gateway and formed its Blueprint Alliance with Amazon Web Services and CrowdStrike (CRWD), and Okta's own Oktane announcement also lists Databricks, Google Cloud, Salesforce and ServiceNow as members. SiliconANGLE also pointed to the acquisition of CyberArk by Palo Alto Networks (PANW) as a sign of consolidation in identity security.
Thiemann described the agent market as something of a Wild West with a lot of change under way, according to SiliconANGLE. SiliconANGLE's own read was that this complexity favors layering several defenses over relying on any one platform to dominate.
What’s next
Neither RSA nor Radware disclosed pricing, and RSA's air-gapped version is not due until 2027. Before committing, security teams can use existing tools to count the agents and MCP servers already in use, then test shortlisted vendors on whether they find unsanctioned agents, enforce policy at runtime and produce evidence an auditor would accept. Protocols such as Cross App Access, which Okta describes as open, are worth tracking, since interoperability will decide how easily a point tool can be swapped out later.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
