AI & Security Trends and Predictions for 2025 by Itamar Golan, Co-founder and CEO of Prompt Security and Core Member of OWASP Top 10 for LLMs
It’s that time of the year again: Predictions season!
Yes, we know that predicting the future is almost impossible, but since last year we got some of our predictions right, we’ve emboldened ourselves to do it again.
But hey, at least we’re being honest with ourselves. Don’t miss out this video reviewing what were the hits and misses of Itamar’s predictions of last year:
So what do we believe 2025 has in store for us in the AI & Security space?
Prompt Security’s AI & Security Predictions for 2025
The Future of Work with AI
Contrary to widespread concerns, I don’t expect AI to eliminate jobs in 2025. Instead, it will serve as a powerful tool to enhance human capabilities. Agentic AI systems will work alongside humans, like in customer service, sales outreach, marketing content creation, software development and healthcare applications, among others. This means that very soon, 30% of our tedious and repetitive tasks will be automated, giving us more time to focus on creative, innovative and interesting pursuits.
I believe we will also see a significant shift as the multi-modality of AI becomes more mainstream (video, audio, etc.), as opposed to the majority of the use of AI which has been text-based. This creates new opportunities for human-AI collaboration.
Organizational AI Adoption
The democratization of LLM access, driven by ever-decreasing prices, is enabling broader adoption across organizations. Additionally, specialized AI solutions will increasingly be moving away from OpenAI's dominance, with alternatives like Claude gaining traction in specific domains such as coding, which is something we’re already starting to see.
Security Landscape
The security landscape is evolving rapidly with challenges on several fronts.
On the one hand we have the rise of AI-powered malware and cyber attacks, which is particularly concerning, as AI is making malware creation accessible to individuals with basic technical knowledge. This democratization of cyber threats poses new challenges for security teams. Gartner predicts that by 2027, around 20% of cyberattacks will be AI-related.
On the other hand, as organizations increasingly embed AI into their sensitive workflows, especially given the expansion of Agentic AI, the security of AI systems themselves becomes a critical concern. By the end of 2025, I would venture to say that at least one in four Fortune 500 companies will have implemented dedicated AI security solutions.
And lastly, in terms of using AI to improve security workflows, I expect this to continue growing - while this is not new - alert triaging, handling, escalations and remediations will continue to see more and more use of AI. This is already a market on its own.
Regulatory Environment
The regulatory landscape for AI is developing along divergent paths globally. The European Union is taking a risk-based approach about AI through the EU AI Act, implementing comprehensive regulatory frameworks. In contrast, I expect the United States to adopt a more permissive approach under potential libertarian economic policies, allowing for greater flexibility in AI development and deployment, emphasizing national security and economic competition with China.
Agentic AI
AI chatbots use generative AI to provide responses based on a single interaction. A person makes a query, and the chatbot uses natural language processing to reply.
In my opinion, the next frontier of artificial intelligence will be agentic AI, which employs sophisticated reasoning and iterative planning to autonomously solve complex, multi-step problems. It is poised to enhance productivity and operations across various industries.
Agentic AI systems process vast amounts of data from multiple sources to independently analyze challenges, develop strategies, and execute tasks such as supply chain optimization, cybersecurity vulnerability analysis, and assisting doctors with time-consuming tasks.
I believe that by 2025, we will see a significant increase in resources shifting from single-interaction procedures with LLMs to this multi-step approach of agentic AI, which will gradually solve complex problems for us autonomously.
Be sure that in 12 months we'll be back to analyze what were our hits and misses. In the meantime, here's to a joyful 2025 filled with safe & secure adoption of AI.
