Key Insights
The IBM Cost of a Data Breach Report 2026 puts a staggering price tag on the rise of AI-powered cybercrime.
Malicious AI-driven breaches now cost organizations an average of $6.04 million, about $1 million more than malicious breaches without AI. At the same time, AI-driven attacks increased 56% year over year, with deepfake and impersonation attacks representing the largest share of incidents.
But perhaps the most important finding is what AI is being used to amplify. Phishing remains the leading initial attack vector for the fourth year running. Social engineering and abuse of valid accounts are still among the costliest ways into an organization. AI hasn’t displaced these human-centric threats; it is making them dramatically easier to produce, personalize, and execute.
When attackers can automate the work required to imitate trusted people, compromise legitimate identities, and manufacture convincing interactions, defenses built around static indicators become increasingly easy to outmaneuver. The advantage shifts toward systems that can understand how people and identities normally behave—and recognize when something is off.
Below, we break down the IBM findings that matter most for defenders confronting this new generation of AI-powered, human-centric attacks.
AI Is Making Human-Centric Attacks More Expensive
IBM found that more than one in four organizations experiencing a malicious attack faced one that was AI-driven, a 56% increase from the previous year. Those incidents were also substantially more expensive: malicious AI-driven breaches averaged $6.04 million, compared with $5.03 million for malicious attacks without AI.
The composition of those attacks matters. AI-generated phishing and other communications represented 17% of AI-driven attacks. IBM attributes the trend in part to generative AI making social engineering cheaper to create and harder to detect.
That tracks with the broader breach data. Phishing was once again the leading initial attack vector, while social engineering attacks such as help desk impersonation and MFA fatigue averaged $5.23 million in breach costs. Abuse of valid accounts followed closely at $5.07 million.
AI is removing the practical limits from attacks that exploit people, identities, and trusted relationships. What once required significant manual effort can now be produced far more efficiently, giving attackers more opportunities to find the one employee, vendor, or account that opens the door.
AI Is Also Creating a Measurable Defensive Advantage
Fortunately, IBM’s data shows the reverse is also true: organizations that put AI to work defensively are seeing materially better outcomes.
Organizations using security AI and automation extensively experienced average breach costs of $4 million, compared with $5.93 million among organizations using none. They also identified and contained incidents in 215 days, 65 days faster than non-adopters.
That nearly $2 million gap is difficult to dismiss. But the more interesting question is where AI creates the most value.
IBM found that adoption remains weighted toward detection and investigation, while prevention continues to lag. Only 33% of breached organizations reported extensive use of AI and automation for prevention, compared with 41% for detection and 39% for investigation.
For human-centric attacks, waiting until investigation is already underway gives the attacker too much room. A convincing impersonation attempt, compromised account, or fraudulent request can create significant damage without ever delivering malware or triggering a traditional indicator. Prevention depends on recognizing the behavioral anomaly before the interaction succeeds.
Trusted Identities and Relationships Remain High-Cost Attack Paths
The same pattern shows up beyond phishing. Supply chain compromise was the second most common initial attack vector in IBM’s study, while abuse of valid accounts remained one of the most expensive, with average breach costs of $5.07 million.
These attacks are difficult to stop for a simple reason: they begin with something the organization already trusts.
A compromised vendor, legitimate employee account, or familiar business process gives attackers cover that malware and obviously malicious infrastructure cannot. By the time suspicious activity appears, the attacker may already be operating inside a relationship or identity that security controls have been conditioned to accept.
That trust also makes these incidents harder to unwind. Supply chain compromises took an average of 258 days to identify and contain—among the longest timelines in the report—while social engineering incidents took 254 days and phishing attacks took 251.
AI makes that problem more acute. Generative models can help replicate tone, construct plausible requests, and adapt messaging instantly, making abuse of a trusted identity far more difficult to distinguish from normal activity.
The Cost of a Breach Is Still Measured in Time
For all the changes AI is bringing to cyberattacks, one rule from last year’s report still holds: the longer an attacker remains undetected, the more expensive the breach becomes.
IBM found that the average time to identify and contain a breach increased to 247 days this year, reversing a five-year decline. Breaches lasting longer than 200 days cost an average of $5.65 million, compared with $4.32 million for incidents resolved within 200 days.
Who discovers the breach matters too. Internal IT and security teams identified and contained incidents in 209 days on average, 15% faster than the global average. When attackers themselves disclosed the breach, that timeline stretched to 268 days. Third-party discovery took even longer, at 281 days.
This is where the defensive AI findings become especially important. Organizations using AI and automation extensively cut identification and containment time by 65 days compared with organizations using none.
Modern Defense Starts With Behavior
At Abnormal, the IBM Cost of a Data Breach Report 2026 reinforces a principle that has become increasingly important as attackers adopt AI: when malicious activity is designed to exploit normal human interactions, security needs to understand those interactions at a behavioral level.
Traditional detection approaches were never designed for an environment where attackers can rapidly generate convincing communications, mimic trusted identities, and adapt attacks without relying on malicious payloads. AI accelerates that shift because it makes the human side of cybercrime dramatically more efficient.
Behavioral AI gives defenders a way to counter that advantage. By understanding normal patterns across employees, vendors, identities, relationships, and communications, security systems can identify the subtle deviations that reveal an attack even when the message, sender, or request appears legitimate.
That approach underpins how Abnormal protects cloud email, identity, and collaborative environments like Microsoft 365. Rather than relying primarily on static indicators, the platform uses behavioral context to identify socially engineered attacks and compromised accounts before they can turn trusted interactions into costly incidents.
IBM’s findings suggest that this kind of AI-powered defense is becoming increasingly consequential. Organizations that used security AI and automation extensively saved nearly $2 million in average breach costs and responded 65 days faster than organizations that used none. Yet only 36% of breached organizations reported extensive use of these technologies across the security lifecycle.
As AI drives breach costs higher, behavioral intelligence offers a path to recognizing threats sooner, stopping them earlier, and preventing a convincing interaction from becoming a multimillion-dollar incident.
Ready to stop costly human-centric attacks with behavioral AI? Request a demo.
Request a Demo
