CERT-W Report 2026: Incident response in an AI-accelerated threat landscape
Published November 17, 2026
- Cybersecurity
In brief
- Financial gain still drives most attacks (57%), but data is now the lever: stealing and reselling it outweighs encryption.
- Speed is the new battleground: the fastest attackers reach impact in under 24 hours, while intrusions stay hidden for 141 days on average, with a median detection delay of 18 days.
- AI changes the economics of attack: flaws are found and exploited faster than organizations can fix them.
- Vulnerability exploitation becomes the leading way in (36% of incidents, vs. 20% last year).
- Identities are the new perimeter: one stolen account can open mailboxes, SaaS platforms and business data.
Cybersecurity report 2026: what CIOs, CTOs and CISOs need to know
Motivations: attackers follow the money, and the money is in data
Profit remains the primary driver, behind 57% of the incidents we handled (65% in 2025). What has changed is the business model:
- Data is the new ransom: exfiltration and resale of data or access (36% of financially motivated attacks) now outweigh ransomware (32%).
- Reputation as a weapon (13%): tracked for the first time, these attacks publish stolen data in full to damage the victim and build the attacker’s notoriety.
- Espionage stays in the shadows (3%, down from 17% in 2025): these intrusions are inherently difficult to detect and attribute and can persist unnoticed for years or be contained before any visible impact. In our analysis, this does not mean espionage has declined, but that other teams, including state-related entities, handled these cases.
- Motives remain unclear in 27% of cases (vs. 18% in 2025): limited impact, early disruption or insufficient evidence keep some investigations from attributing a clear motive.
This shift shows that data needs stronger protection: locked systems remain an issue, but are no longer the priority.
The organizations that come through a crisis are those that see the attacker early and decide fast.
Entry points: exposed systems and stolen identities overtake phishing
How attackers get in changed markedly in a single year, in line with global trends:
- Vulnerability exploitation (36%, vs. 20%): internet-facing applications, remote access gateways and security appliances. Where the exploited flaw could be identified, nearly three-quarters were already known (N-day) and patchable.
- Stolen or purchased credentials (30%, vs. 22%): attackers simply log in, and their activity looks like business as usual.
- Phishing and social engineering (15%, vs. 38%): email phishing declines in well-equipped organizations, but manipulation is moving to phone calls, text messages and the abuse of genuine login pages.
The perimeter is no longer a network boundary. It is the sum of every exposed system, every identity and every partner connected to the organization.
Key takeaways for leadership teams
- Remote access remains a front door: VPN gateways, web applications and security appliances are among the most targeted systems, through an unpatched flaw or a compromised account.
- Recovery systems are priority targets: attackers increasingly go after identities, hypervisors and backups before exfiltrating data and applying pressure.
- One account can compromise an ecosystem: 30% of incidents involved SaaS, Microsoft 365 or CRM environments; a single compromised mailbox was used to target more than 900 contacts.
Emerging trends in 2026
In 2026, Claude Mythos and subsequent AI models showed that AI can discover and exploit previously unknown vulnerabilities end to end, in minutes rather than weeks. The effects already show in the data: the number of vulnerabilities published has tripled in a year, and those exploited on day one rose by half.
Patching everything, constantly and at speed, isn’t realistic. The organizations that cope best have rebuilt their vulnerability management pipeline and prioritize based on real-world exploitation, which covers around 80% of attacked vulnerabilities for a fraction of the effort.
Attackers increasingly log in rather than break in. Stolen passwords, hijacked sessions and abused login flows give them access that is hard to tell apart from normal use. Device code phishing shows how quickly this spreads: first observed in state-linked operations, it is now commoditized through phishing-as-a-service and widely adopted by cybercriminals.
The financial impact can be direct. In one case, attackers quietly watched a company’s mailbox for four months, then impersonated an employee to change a partner’s bank details: around €50,000 was diverted. In another, a single purchased account on a medical platform was enough to steal patient data and launch mass phishing.
Ransomware and data exfiltration remain the top threats in 2026, but the model is shifting from encryption toward data theft and extortion: among financially motivated attacks, exfiltration and brokering (36%) now outweigh ransomware (32%).
According to Wavestone’s 2026 AI Cyber Benchmark, based on 30 large public and private sector clients, 40% of enterprises develop proprietary AI models and 50% design AI systems internally using existing frameworks or pre-trained models. Adoption often outruns security: rushed procurement, vibe-coded applications and late security reviews; exposed apps and AI gateways linked to models; API keys and broad default permissions; agents that can read, alter or delete business data and can be hijacked through prompt injection.
At the same time, attackers use AI for deception at scale, faster vulnerability research and AI-assisted or agentic malware, with major new cases made public in 2026.
What this means for CIOs, CTOs and CISOs
The full report details eight operational priorities. For leadership teams, they come down to five decisions:
- Manage exposure as a business risk: know what is visible from the internet and fix what is actively exploited first, on a clock measured in days, or even hours for the most critical systems.
- Make identity the new perimeter: phishing-resistant authentication wherever it matters, and close watch on accounts, sessions and connected applications.
- Protect the ability to recover: isolate backups, virtualization and administration systems, and test restoration end to end, industrial systems included.
- Invest in detection and decision speed: keep the logs, hunt for attackers already inside, and rehearse crisis decisions for an attack that unfolds in a day.
- Secure AI and third parties before scaling them: treat AI platforms, agents and supplier access as critical assets, under the same governance as core systems.
Conclusion
2026 confirms that attackers have changed their tactics: they exploit what is exposed, log in with what is stolen, and move faster than most organizations can detect.
AI accelerates every step and puts rapid response capabilities under pressure.
For CIOs, CTOs and CISOs: set a new defensive baseline to reduce exposure, secure identities, protect recovery and rehearse the response, before the crisis arrives. It will be key for 2027 to define the new cyber KPIs, measured in hours rather than days:
- How many hours does it take to spot a new system exposed to the internet?
- How many hours does it take to take an exposed system offline when the risk is critical?
- How many days does it take to patch an actively exploited vulnerability, or to put compensating controls in place?
Setting these targets is not a task for IT alone: top management, CIOs, CTOs and cyber teams need to agree on them together.
About this report
The report is built on the incident reports produced by CERT-Wavestone analysts, consolidated in a single database and put in context by Wavestone’s Cyber Threat Intelligence team. It covers organizations of all sizes, 82% from the private sector and 18% from the public sector. Public administration, manufacturing, construction and real estate, and health were the most represented sectors.
All cases are anonymized. Figures reflect CERT-W investigations, not the overall distribution of incidents.
Look back at last year’s findings in the CERT-W Report 2025.