News
Emanuela Zaccone uncovers the AI Trifecta for security
When you have a chance to pick the brain of Emanuela Zaccone, AI Staff Product Manager at Sysdig, you are in for a treat. At Edgecase 2025, she shared insights from literally spending her days building AI solutions for cybersecurity. The audience was looking forward to hearing what she had to say.
Emanuela explained that modern attackers don’t wait. They’re fast, stealthy, and willing to experiment. The AI-powered security shift has accelerated our arms race as we face scale and unpredictability at machine speed. This year's emergent trend: threat actors exploiting CI/CD pipelines and moving through cloud workloads before you can blink, all with some help from AI.
In her session, Emanuela guided us through the 2025 cloud threat landscape, revealing where attackers are going next and how runtime security has become a non-negotiable pillar of any modern defense.
She also shared insights into the AI security trifecta, offering practical recommendations on how to:
- secure the AI being adopted;
- integrate AI into your security operations;
- and defend against AI-powered threats.
The place of AI Security in the Multiverse Saga
When using the Marvel analogy, AI Security is a different timeline that started in the 2020s. Everyone started using ChatGPT a couple of years ago. Even Emanuela’s parents knew about it, she says. It soon occurred to her that AI is a totally new dimension, and it’s here to stay. AI is a huge universe that will change everything.
Recent Gartner analyst research shows this:
- 58% of organizations have deployed or are piloting GenAI use cases;
- CVEs increased by 34% every year;
- 41% of the cybersecurity community embraces GenAI for threat hunting and modeling.
We’re in the middle of the fastest evolution in cloud computing history. But cloud innovation has outpaced cloud security.
Cybercrime costs 11 times more than AI is worth
AI companies are worth 1 trillion dollars in value. However, cybercrime costs companies 11 trillion dollars in damages. Emanuela delivers this shocking fact to wake every audience member who still needed to be awakened.
Emanuela: “The attack surface has exploded. Attackers already use AI automation, which is constantly and dynamically changing. We need a shift in the approach of security. A new runtime model just for AI is needed. LLMs that are trained with training data by attackers are already very challenging. But the more sophisticated AI attacks get, the harder it is to prevent them. The threats just add up when AI can use reasoning, make its own tools, and has a memory of what works and doesn’t work.”
60% of containers live for one minute or less. The Sysdig Threat Research Team uses runtime analysis to identify threats that statistical analysis won’t find, specifically for this reason.
The three pillars of AI Security
Pillar 1: Securing AI is foundational, not optional
Don’t feed sensitive information into LLMs, for example. Implement guardrails in your configuration (enforce least privilege) and clearly communicate AI usage policies. Continually monitor everything!
Pillar 2: Use AI in security
This is the only way to keep pace with machine-speed threats. Attacks unfold in minutes, with GenAI investigations cut to 3.5 minutes, not days. Meanwhile, the time to respond to cloud security incidents using AI is 76% faster.
AI transforms security ops from bottleneck to force multiplier. Everything is much faster with AI. But for now, keep a human in the loop!
Quickly apply remediation strategies to boost security with minimal effort and risk. For example, with AI-assisted VM remediation strategies.
Pillar 3: The only way to beat AI-powered attacks is with AI-powered defense
Attackers already weaponized AI with phishing and malware evolution. 49% of leaders report improved security posture from AI. Multi-layer runtime detection is key. Defend against evil AI with a multi-layer runtime defense. Fight fire with fire.
Even when attackers use AI-generated malware, they keep human feedback too. Be prepared, use speed to your advantage, and use AI. It pays back, in money (ROI), resilience, and trust.
AI Security is a great investment
88% of early adopters see positive ROI from AI agents. Organizations adopting GemAI have 6-10% revenue growth. 55% cite direct security improvements. Over 95% of low-risk noise reduction with Agentic AI. Prioritize remediation with agentic AI and save up to 80 hours per week.
At the end of Emanuela’s talk, we appear to see a video of an attack, but it turns out to be a challenge for the audience: beat three levels in a security quiz, and they could win Sysdig collectables. A nice way to activate the engineers in the room, after an inspiring session.
