Cybercriminals are utilizing AI to maneuver sooner, and firms are paying a premium when these assaults result in information breaches.
IBM discovered that AI performed a job in a single in 4 malicious breaches examined for its 2026 Price of a Information Breach Report, a 56% enhance from the earlier 12 months. These incidents value organizations a mean of $6 million — about $1 million greater than the worldwide common breach value of $4.99 million.
Carried out by the Ponemon Institute, the research examined breaches skilled by 602 organizations between March 2025 and February 2026.
The findings additionally reveal a rising imbalance for safety groups: Many organizations are utilizing AI to detect assaults already underway, however far fewer are deploying it to search out vulnerabilities earlier than attackers can exploit them.
“What’s altering is the economics of cyberattacks. AI is making assaults sooner and cheaper, whereas breaches maintain getting dearer,” stated Suja Viswesan, vp of IBM Safety Software program. “When organizations have an prolonged hole between discovery and remediation, that imbalance exhibits up immediately in breach prices.”
Corporations are utilizing AI, however not the place attackers are transferring quickest
Organizations that used AI and automation extensively in safety operations reported breach prices averaging $1.93 million lower than organizations that didn’t use the applied sciences, in line with IBM.
Nonetheless, IBM discovered that many corporations are nonetheless utilizing AI primarily after assaults start. Greater than half of breached organizations reported utilizing AI brokers for menace detection and containment, however solely 18% used brokers for vulnerability administration.
That hole may turn into more and more consequential as attackers use AI to speed up vulnerability discovery and exploitation.
IBM discovered that greater than 20% of organizations skilled breaches involving AI fashions or functions. The commonest causes have been weaknesses round AI programs reasonably than the fashions themselves, together with compromised APIs, functions or plug-ins, in addition to cloud misconfigurations affecting AI workloads.
Essential industries face increased AI-related dangers
Essential infrastructure organizations skilled the vast majority of reported AI-driven assaults, accounting for 62% of incidents. Monetary companies and vitality corporations noticed a number of the highest concentrations of AI-related assaults.
Monetary companies breaches averaged $6.3 million, whereas vitality breaches averaged $5.2 million, highlighting the potential for cyber incidents to have an effect on important companies, provide chains and broader economies. Healthcare remained the most costly business for breaches for the thirteenth consecutive 12 months, with common prices reaching about $6.6 million, in line with IBM’s findings.
Should-read safety protection
AI governance stays a weak level
Whereas corporations are growing AI adoption, many are struggling to manage how the know-how is used internally. Shadow AI — using AI instruments or programs with out organizational approval or oversight — was concerned in 43% of the incidents IBM studied, greater than twice the earlier 12 months’s share.
IBM additionally discovered that 92% of organizations affected by AI-related breaches lacked correct entry controls for his or her AI programs. The discovering means that weak id administration, restricted governance and poor oversight could also be leaving AI programs unnecessarily uncovered.
The report additionally recognized gaps in primary information safety. Solely 37% of breached organizations stated they encrypted delicate information each at relaxation and in transit, whereas many lacked a whole stock of encryption keys, certificates and different cryptographic belongings.
Safety groups face a race towards AI-powered attackers
The report suggests that companies can not rely solely on conventional safety enhancements. AI is rushing up assaults, however it could possibly additionally assist organizations reply sooner if deployed appropriately.
The problem is deciding the place AI must be utilized. Many corporations are utilizing AI to research alerts after suspicious exercise seems, however fewer are utilizing it to proactively determine vulnerabilities earlier than attackers discover them.
Additionally learn: The Largest Information Breaches of 2026, Ranked by Influence.
