F5 CEO François Locoh-Donou explains how leading enterprises can embrace AI with confidence by recognizing it as the most vulnerable technology ever deployed at scale.
Nearly every leader I speak to feels pressure to move faster on AI. In the push to find efficiencies and innovate, security is too often overlooked.
Frontier AI has broken the traditional security timetable. The Zero Day Clock project, a top cybersecurity body of research, found that in 2018, the median time from a disclosed security vulnerability to exploitation was more than two years. By 2024, that timeframe had collapsed to four hours. Today, advanced models can identify and weaponize flaws before they are publicly disclosed, creating a negative time window between disclosure and exploitation. Meanwhile, the average organization still takes 20 days to test and deploy a patch. Let that sink in.
Make no mistake: I am an AI optimist. At F5, we have more than 23,000 customers in 170 countries, and we’re partnering with many to scale their AI use cases. Consider these organizations: an elite professional sports league deploying internal AI apps to analyze player health and team strategies; a global bank with 50 AI use cases ready to launch; and an enterprise workforce technology provider running dozens of live AI applications, with hundreds more on its roadmap.
These are real-life examples. Like many leading organizations, at these businesses, AI is no longer a collection of trial-and-error chatbots or isolated productivity hacks but is core to running the business.
The key to scaling AI successfully is understanding that it is the most vulnerable technology ever deployed at scale. Leaders must acknowledge that and prepare accordingly. Traditional security is insufficient. Gone are the days of just protecting against bad actors and fortifying critical weak spots.
The rise of shadow AI is a prime example. A Gartner survey of cybersecurity leaders between March and May of 2025 found that 69% of organizations suspect or have evidence that employees are using prohibited public AI, creating ungoverned and unmonitored data flows. That percentage is likely higher now. When well-meaning employees paste proprietary data into these unvetted tools, the business’s risk of exposure grows.
AI models themselves are also risky. Beyond traditional data leakage, even models that are approved by IT departments are exposed to new attack vectors such as prompt injection, model extraction, and jailbreaking. As AI adoption and the use of autonomous agents expand by the billions, enterprise attacks will explode.
As a CEO and board member, I know security teams are often seen as a brake on business velocity and an obstacle to shipping new innovations quickly—sometimes known as the office of no. In the AI era, that mindset is fatal. When a single breach can dismantle customer trust, trigger regulatory sanctions, and erase enterprise value overnight, robust security is mission critical.
Executive leaders must have an active, strategic partnership with their chief information security officers (CISOs) and their teams. The best CISOs act as a map to the maze of AI security, recognizing that traditional security tools and fragmented point solutions are inadequate in the AI era. Securing only portions of the problem simply creates gaps and blind spots easily exploitable by today’s—and tomorrow’s—AI threats.
The solution is an end-to-end, virtuous-loop AI security platform:
- Continuous discovery: Map the entire enterprise AI footprint, uncovering shadow tools and classifying the intent behind every user and agent interaction, without disrupting employee productivity.
- Automated offensive testing: Conduct continuous AI red teaming to stress-test AI models, applications, and agents against tens of thousands of dynamic attack scenarios to find the cracks before attackers do.
- Remediation and runtime protection: Eliminate the guesswork by automatically translating red-team vulnerability reports into active guardrails. Rather than burdening stretched security teams with manual triage, automated remediation seals the cracks at runtime, blocking prompt abuse and preventing data exfiltration in real time.
This lifecycle approach is precisely why pioneering organizations are succeeding with AI—and why the CISOs doing it right can become the office of yes!
The three examples I highlighted earlier are real, and they are demonstrating this today: The sports league deployed on-premises red teaming and guardrails to safeguard sensitive team and player data. The global bank has the secure infrastructure it needs to maintain peak protection while supporting an expanding portfolio of AI applications. The workforce technology provider gained the visibility and control it needed to scale its AI roadmap safely.
At F5, we have spent three decades sitting in the live data path of the world’s most critical applications, giving us unmatched historical perspective and a unique advantage when it comes to securing AI.
AI security is not a one-time compliance milestone—it is a permanent operating rhythm that empowers organizations to innovate, protect enterprise trust, and lead the future. The time to talk to your CISO to map your AI security strategy is now.
Note: This content was created by F5.