Cybersecurity used to be simple: use strong passwords, run regular scans, update your anti-virus software, watch out for ‘dodgy’ emails, and never open a suspicious link or attachment. But now?
AI has changed the rules and is reshaping the battlefield. Today, access to this technology helps to strengthen defences and monitor threats in real-time, but it also enables faster and smarter attacks.
Reacting is no longer a defence option. Cybersecurity is now about real-time proactivity; monitoring, predicting, and adapting defences to stay one step ahead of potential threats that are also constantly evolving.
How is AI reshaping cybersecurity?
AI is quietly transforming how cybersecurity actually works behind the scenes. What used to take teams of analysts hours, or days, to investigate, machine learning systems now process in seconds. AI helps with the boring but critical work; filtering alerts, isolating infected devices, and even auto-patching vulnerabilities, and that is a big deal when security teams are stretched thin. But it is not just about speed, it is about finding the weird stuff: subtle patterns or unusual behaviours that humans would easily miss.
Is it perfect? No. But it has become a crucial extra set of eyes, fast, tireless, and always learning.
The EIT Deep Tech Talent Initiative’s Pledgers’ Share & Connect in May focused on the work some of our Pledgers are carrying out in this arena. One presentation from Tiberiu Baraboi, Solutions Architect and Founder of Expertware, introduced SIEMBIOT, Expertware’s collaborative and EU co-funded, cybersecurity research platform, built on top of industrialised SOCaaS, that focuses on threat intel dissemination and provides an educational tool in the field of cybersecurity, enhancing existing teaching methods by offering a virtual laboratory based on real and anonymised data.
Key AI-driven cyber threats in 2025
These days, cyber criminals don’t all have to be tech geniuses. With AI being more accessible, it makes cyber-attacks easier to execute and harder to detect because they are more convincing.
- Phishing: Forget spam, phishing attacks are now social engineering, at scale! Gone are the clunky emails and sketchy links, and replaced with AI-written ‘almost perfect’ emails that mimic the ‘purported’ company’s tone and language, and personalised using scraped social media data. The only giveaway is the email address!
- Malware: With AI-powered ransomware, hackers are developing self-learning code that can adapt and evolve mid-attack to dodge detection tools, shift tactics, and re-encrypt if blocked.
- Deepfakes: These are no longer an internet curiosity confined to switching famous actors into different movies, even Disney has improved their visual effects using high-resolution deepfake face swapping technology to de-age characters or revive deceased actors, as they did for the characters of Princess Leia and Grand Moff Tarkin in Rogue One. But in the cybersecurity world, scammers are using deepfakes to create AI visuals and voice cloning for fraudulent purposes.
These active threats are growing more sophisticated every day, and they’re not limited to big corporations. Anyone can be a target when AI lowers the barrier to entry for cybercrime.
Effective AI-powered cyber defence tactics
AI might be helping attackers, but it is also giving defenders some serious firepower. One of the most effective tools? Machine learning models that can spot weird patterns in huge piles of data, fast. Not just known threats, but strange behaviours that suggest something is not right, like a login at 3 a.m. from an unexpected location or a system accessing files it never touches.
Still, AI is not magic. It makes mistakes. Analysts can spot context that an algorithm just cannot, like whether that 3 a.m. login was legitimate, or someone simply forgot to log off, which is why human oversight matters more than ever. It is the combination of machine speed and human judgement that works.
There is also the job of protecting the AI itself, because if hackers mess with training data or exploit model vulnerabilities, they can poison results or sneak through undetected. So, defending your defences and keeping up with the latest developments in AI is now part of the job.
EIT Deep Tech Talent Initiative Course Catalogue
As of July 2025, the EIT Deep Tech Talent Initiative has supported the skilling and upskilling of 1 million talents across Europe over the past 2 years. With over 200 courses and training programmes, 86 (38%) are dedicated to Artificial Intelligence & Machine Learning (including big data), and 35 (15.5%) are focused on cybersecurity and data protection.
These courses offer a great opportunity for European talent to skill, upskill, or reskill in these in-demand sectors.
Partners
