Senior Research Engineer

GISEC 2026: Tenable to highlight Agentic AI-driven exposure management. At the event, Tenable will showcase its Tenable One Exposure Management Platform. Tenable announced its participation at GISEC Global 2026, taking place at the Dubai Exhibition Centre (DEC), Expo City Dubai from 16-18 September. At the event, Tenable will showcase its Tenable One Exposure Management Platform, which unifies visibility across an organisation's entire attack surface. Building on this foundation, the company will highlight how its AI capabilities, including Hexa AI and AI Exposure, help close critical security gaps before attackers can exploit them. As organisations across the Middle East continue their digital transformation journey and integrate generative AI into enterprise operations, security teams are facing an influx of expanding attack surfaces and rapid threat cycles. At GISEC Global 2026, Tenable will demonstrate how these AI capabilities bridge the gap between exposure discovery and automated remediation. "AI and cloud adoption are moving faster across the Middle East than most organisations' security controls can keep pace with," said Maher Jadallah, Vice President, Middle East and Africa at Tenable. "Attackers are already using AI to find vulnerabilities faster than defenders can patch them. As initiatives like the UAE National Cybersecurity Strategy push organisations toward more resilient infrastructure, security teams need a clear, prioritised view of where they're exposed, not another disconnected tool." Visitors to GISEC can schedule briefings with Tenable security experts and experience live demonstrations of the Tenable One platform, including Tenable Hexa AI and AI Exposure, throughout the event at booth number H7-B130 in Hall 7. Tenable One brings together two distinct AI capabilities. Tenable AI Exposure helps organisations discover, assess and secure how AI is being used across their environments. Tenable Hexa AI is the platform's agentic engine, using AI to coordinate agents, automate security tasks and accelerate remediation. Put simply, AI Exposure helps organisations secure their use of AI, while Hexa helps them use AI to improve security operations. Together, they advance Tenable's preemptive security strategy by helping organisations reduce AI-related risk and act on cyber exposure more efficiently.

Tenable & OpenAI launch AI inspector for cyber agents. Tue, 8th Sep 2026 (Today) Tenable is working with OpenAI to launch CyberAgents Exchange AI Inspector, a review process intended to assess AI agents and related components before deployment. The move reflects a joint effort by the cybersecurity company and OpenAI as businesses increase their use of agentic AI tools in enterprise environments. CyberAgents Exchange AI Inspector is designed to review AI agents, skills, MCP servers and multi-agent playbooks listed on the CyberAgents Exchange, operated by Tenable. The process combines OpenAI GPT cyber models, inspection through Tenable One AI Exposure and review by Tenable researchers. The aim is to help security teams identify and prioritise risks before community-built AI components are introduced into corporate systems. Tenable framed the initiative as a response to a growing market for shared AI tools built for cyber defence use cases. The collaboration grew out of Tenable's involvement in the OpenAI Daybreak Defence Network, a programme that brings cybersecurity companies into OpenAI's ecosystem. CyberAgents Exchange AI Inspector is expected to become available in September. The exchange The CyberAgents Exchange launched as an open-source registry focused on cybersecurity applications for AI agents and related tools. It now contains more than 100 community-submitted AI components following a recent SWARM build event hosted by Tenable. That growth points to a broader shift in the security market, as vendors and practitioner communities experiment with AI systems that can carry out multi-step tasks, connect to external services and coordinate with other agents. For security teams, the challenge is that such tools can also introduce new risks through software flaws, unsafe permissions, data handling issues or weak oversight. By introducing a formal inspection process, Tenable is seeking to add more structure to how those components are assessed before use. Its existing AI exposure technology is intended to support that process by examining risks associated with AI systems and their surrounding infrastructure. Security concerns The announcement comes as companies face pressure to adopt generative and agent-based AI while also addressing concerns from security and governance teams. In practice, that has created demand for products and review methods that can test not only models, but also the tools, workflows and connectors built around them. MCP servers, one of the component types covered by the new process, connect AI systems with outside resources and applications. Multi-agent playbooks are structured workflows in which several AI agents perform separate roles to complete a larger task. Security specialists have increasingly warned that these systems may expand an organisation's attack surface if deployed without proper vetting. Community-built components can broaden access to innovation, but they can also create uncertainty around provenance, permissions and operational behaviour. Eric Doerr, Chief Product Officer at Tenable, said the collaboration is intended to address that issue. "Agentic AI will only reach its potential in the enterprise if security teams can trust the components being introduced into their environments," Doerr said. "By combining OpenAI GPT cyber models with Tenable's security expertise and researcher review, we're building a more rigorous way to inspect community-built AI components before they're used in enterprise environments. This is an important step in applying frontier AI to help defenders identify and prioritise risk across the growing agent ecosystem." OpenAI's cyber-focused summit, where the initiative was introduced, brought together security industry leaders to discuss defensive uses of advanced AI systems. The discussions focused on how AI reasoning tools can be integrated into existing security products and workflows to help defenders handle threat findings and remediation more quickly. For Tenable, the tie-up also extends its effort to position itself in the emerging market for AI security oversight. As organisations test agent-based systems for security operations and other tasks, vendors are competing to offer assessment tools that can give risk teams clearer visibility into what those systems do and how they connect to enterprise environments.

Tenable builds an openai-powered vetting gate for MCP servers. Tenable and OpenAI are building Exchange Inspector, a security review for AI agents, skills and MCP servers on the CyberAgents Exchange. Here's what it means. Key takeaways. * 1Tenable and OpenAI unveiled the CyberAgents Exchange AI Inspector on September 3, 2026 at OpenAI's Intelligence at Work: Cyber Summit; Tenable says it is expected to be available in September. * 2The review combines OpenAI GPT cyber models, skills inspection via Tenable One AI Exposure, and human review by Tenable researchers - a three-layer gate rather than an automated scan. * 3Independent research suggests the underlying problem is real: an Endor Labs analysis of 2,614 MCP implementations found 82% used file operations prone to path traversal, though scanner methodologies vary widely and headline 'percent vulnerable' figures should be read with care. The CyberAgents Exchange launched in August 2026 with a deliberately open pitch: free to use, open to all, no gates. Six weeks later, Tenable is adding one. On September 3, at OpenAI's Intelligence at Work: Cyber Summit, Tenable announced it is collaborating with OpenAI to build the CyberAgents Exchange AI Inspector - a security review process for the AI agents, skills, MCP servers and multi-agent playbooks listed on its registry. The exchange, launched in August 2026 as an open-source, cybersecurity-native directory, now carries more than 100 community-submitted components following Tenable's SWARM build event at Black Hat USA 2026. Tenable says Exchange Inspector is expected to be available in September. The tension is the story. A registry that grows by accepting whatever practitioners contribute eventually has to answer a question enterprise buyers always ask: which of these is safe to plug into production? Tenable's answer is a vetting layer built on frontier models - and it arrives at a moment when the MCP ecosystem's security record is under sustained scrutiny. Three layers: frontier model, platform scan, human researcher. According to Tenable's announcement, Exchange Inspector combines three things: frontier assessment using OpenAI GPT cyber models, skills inspection powered by Tenable One AI Exposure, and expert review from Tenable researchers. The collaboration grew out of Tenable's participation in the OpenAI Daybreak Defense Network, the partner program through which OpenAI says more than 35 enterprise products and partner-operated services are bringing its cyber-tuned models into defender workflows. The architecture matters more than the branding. Static scanners have struggled badly with this class of artifact - one independent audit cited in 2026 security roundups found roughly a 78% false-positive rate from YARA-based MCP scanners, and Endor Labs has noted that conventional SAST tooling does not detect LLM-specific issues at all. A malicious instruction hidden in a tool description is not a code smell; it is prose that reads as documentation. Keeping a human researcher in the loop, rather than shipping a pure automated verdict, is a reasonable read of where the tooling actually is. > "Agentic AI will only reach its potential in the enterprise if security teams can trust the components being introduced into their environments," said Eric Doerr, chief product officer at Tenable, adding that the goal is a more rigorous way to inspect community-built AI components before they are used in enterprise environments. Why a vetting gate, and why now. The independent research on MCP security is uncomfortable reading. Endor Labs' analysis of 2,614 MCP implementations found that 82% used file operations prone to path traversal, 67% used APIs related to code injection, and 34% used APIs susceptible to command injection. BlueRock Security analysed over 7,000 MCP servers and reported 36.7% as potentially vulnerable to server-side request forgery. Censys counted 12,520 internet-accessible MCP services, most of them unauthenticated. The incident record is not theoretical either. CVE-2025-6514 in mcp-remote - a CVSS 9.6 OS command injection disclosed by JFrog in July 2025, in a package with more than 437,000 downloads - demonstrated remote code execution on a client machine simply from connecting to an untrusted remote server. In June 2026, the Miasma worm campaign planted adversarial MCP configuration files across 73 GitHub repositories, executing credential-harvesting payloads against developers who opened them in IDEs that auto-load project-defined servers. One caveat worth carrying: these figures come from different vendors using different methodologies, and "potentially vulnerable" is doing real work in several of them. The direction of travel is clear; the precise percentages are not directly comparable. What enterprise buyers should watch. For security leaders, the useful question is not whether Tenable's gate exists but what a pass actually certifies. The announcement describes a review process for components "available through" the exchange, and Tenable's own X post referred to a review for select components - which suggests coverage will be partial at launch rather than a blanket sweep of all 100-plus listings. Buyers should ask what fraction of the registry is inspected, how often re-inspection happens, and what the verdict format looks like. Re-inspection is the load-bearing detail. The rug pull attack class - where a server passes review and later silently redefines its tools - makes one-time approval structurally insufficient. So does the fact that a server's tool descriptions enter the model's context window with instruction-level authority the moment it is connected, before any tool is deliberately called. A vetting badge that reflects a single point in time is a weaker signal than it looks. Tenable, which reports over 40,000 customers, is not alone here: Palo Alto Networks, SentinelOne, Sophos, Proofpoint and Trend Micro all announced Daybreak-based capabilities the same day. What distinguishes Tenable's move is the target - not the enterprise's own agents, but the open ecosystem of components those agents pull in. That is the harder surface, and the one nobody currently owns. #Tenable #MCP servers #AI agent security #OpenAI #agentic AI #exposure management #supply chain security #CyberAgents Exchange