Top 5 Takeaways for CISOs from RSAC 2025:

AI Security in Focus

By Mike Pearl - Chief Commercial Officer, Javelin AI

The RSA Conference 2025 brought together over 44,000 cybersecurity professionals in San Francisco, establishing itself as the definitive forum for discussions on the intersection of artificial intelligence and cybersecurity. With AI dominating virtually every booth, session, and keynote, this year's conference sent a clear message: AI is reshaping the security landscape at unprecedented speed, and CISOs must adapt or risk falling behind. Here are five crucial takeaways for Chief Information Security Officers navigating this new frontier.

The Rise of Agentic AI Creates a New Security Paradigm

Approximately 75% of organizations are concerned about agentic AI risks. These AI agents, designed to operate with minimal human supervision, introduce an entirely new category of security challenges. "The massive shift in cybersecurity that will be required in the coming years to respond to the arrival of software powered by GenAI and agentic AI can't be overstated," noted one RSA keynote speaker. At Javelin, we leverage leading agent standards, including Langchain’s Agent Protocol, Anthropic’s Model Context Protocol (MCP), Google’s A2A Protocol, and Agntcy OASF and ACP, enabling organizations to enforce security controls, monitor behavior, and mitigate risks throughout the agent lifecycle.

Identity Has Become the New Security Perimeter

With identity-based attacks accounting for approximately 80% of breaches, according to data presented at the conference, identity security emerged as the second major focus area.

"The browser is becoming the new endpoint," noted a security leader at Blue Mantis, highlighting how remote work and SaaS adoption have pushed sensitive data into browser sessions. Javelin Guard, our platform API that works with browsers and other end-user interfaces, extends our platform to areas where AI risk is often overlooked.

AI Security Requires a Full-Lifecycle Approach

According to Cisco's forthcoming 2025 Cybersecurity Readiness Index cited at the conference, 86% of organizations have encountered AI-related security incidents in the past year, including training data poisoning (58%), supply chain compromises (68%), prompt injection attacks, model extraction and theft, and unauthorized model access. "The AI models themselves are inherently unpredictable and non-deterministic," explained a Cisco executive at RSA.

This theme was evident throughout the week, and why Javelin’s Platform provides our enterprise customers with a 360° Closed Loop solution, helping them scale safely by including security from design and development through operation.

AI Is Transforming Security Operations

RSA presentations highlighted impressive performance improvements, including 75% faster threat identification, 60% reduction in incident response time, 50% increase in SOC efficiency, 70% more comprehensive risk assessment coverage, and 65% improvement in automated threat blocking. "While AI can significantly accelerate data analysis and help detect threats faster, it still lacks the human judgment, context awareness, and ethical reasoning our analog colleagues possess," cautioned one panelist.

Open-Source AI Security Collaboration Is Accelerating

Several major announcements highlighted this trend, including Cisco's introduction of Foundation AI, an 8-billion parameter, open-weight reasoning model specifically designed for cybersecurity. "We decided to open source the Foundation AI Security model because, in cybersecurity, the real enemy is the adversary trying to exploit our systems," explained Cisco's leadership.

The development of models like this by leading security experts is essential and core to Javelin’s research and development efforts, which include teaming with and enhancing leading open source models to address emerging threats for our customers.

A Look Ahead

Agentic AI is evolving at breakneck speed, and standards are evolving to help organizations take advantage of the opportunities that AI Agents provide. Industry standards on interoperability, including OSAF, MCP, ACP, and A2A, among others, will continue to evolve. At Javelin, our focus is on embracing open standards that allow our customers to engage with standards in a way that builds security from the beginning and remains flexible to evolve with the fast evolution of Agentic AI. 

RSAC 2025 made it abundantly clear that AI is simultaneously the biggest opportunity and challenge facing cybersecurity leaders today. For CISOs, success will depend on embracing AI's transformative potential while implementing rigorous governance, collaborating across industry boundaries, and maintaining a focus on fundamental security principles.

The conference's overarching message was that despite AI's complexity, core security practices remain essential: visibility into digital assets, strong identity controls, secure development practices, and defense-in-depth strategies. By adapting these principles to the unique requirements of AI systems, CISOs can navigate this rapidly evolving landscape and maintain security effectiveness.

See how Javelin can help you:  www.getjavelin.com