The end of the year is near, and the usual 2026 predictions are expected to hit the press soon. We at KuppingerCole, however, are not huge fans of that genre. After all, hardly anyone goes back and checks how last year’s predictions actually performed. So instead of inventing new buzzwords, let’s look at what we really saw in 2025 and ask a more useful question: will those same threats just get worse in 2026, or should we brace for something genuinely new?
Spoiler: it’s both. The patterns are familiar, but they’re becoming faster, more automated, and more systemic. The “surprises” are less about brand‑new attack types and more about where old techniques show up and how much collateral damage they cause.
What 2025 taught us
If you look across our 2025 research, three themes keep coming back: identity, data, and AI‑driven automation.
On the data side, we’ve been very clear that security has become a platform problem: multi‑cloud, SaaS, and AI workloads have scattered sensitive information across hundreds of services and regions, which is why we keep pushing consolidated approaches in our research. This year’s Leadership Compass on Data Security Platforms underscored exactly that point by evaluating vendors that unify discovery, protection, and monitoring across hybrid environments. “From Perimeter to Persona: Why Data Security Now Starts with Identity” is a whitepaper that explores why modern data protection strategies must integrate tightly with the identity layer to remain effective.
And it’s no longer just about people. In 2025 we finally started treating non‑human identities – workloads, APIs, service accounts, containers, bots, and now AI agents – as a first‑class risk category, with dedicated work like the Leadership Compass on Non‑Human Identities and the blog “Mastering Non‑Human Identity Governance”, to say nothing about the numerous heated discussions about identity in an AI world at the European Identity and Cloud Conference 2025.
Finally, we’ve spent a lot of time this year on AI‑driven threats. For example, our whitepaper “Navigating the Future of Authentication in the Age of AI and Deepfakes” highlights how synthetic media is rapidly eroding the reliability of traditional authentication methods, while the blog “AI Takes the Lead: Identifying Cybercrime Masterminds with Behavioral Analytics” examines how adversaries already exploit advanced models to obscure attribution and accelerate attacks. We also looked ahead in “Will AI Agents Be the Foundation of the Next Security Revolution?”, discussing both the defensive potential of autonomous agents and the new risks they introduce.
These three trends create specific problems when they overlap. AI agents need credentials to work. Those credentials are non-human identities that most organizations can't track properly. When attackers automate reconnaissance, they find these orphaned service accounts faster than security teams can inventory them. Data moving between SaaS platforms crosses security boundaries that assumed humans were making the decisions.
So, 2026 won't just be "more ransomware." Attacks will target the gaps: service accounts nobody knows about, API keys buried in agent configurations, data syncing between platforms with different security controls. Defenders treating identity, data protection, and AI as separate domains will get hit by attacks that chain all three together.
What continues in 2026
Ransomware and data leaks are not going away. If anything, they’ve become the operating system of cybercrime. The new piece is the level of automation and AI usage behind the scenes. Recent studies suggest that around 80% of observed ransomware operations are already using AI for reconnaissance, phishing, malware generation, and evasion.
At the same time, Microsoft’s latest Digital Defense Report shows nation‑state actors adopting AI at scale to generate content, speed up intrusions, and blend cyber operations with disinformation campaigns. So, in 2026, we’re not suddenly facing “new” ransomware but industrialized campaigns that can find soft targets and craft credible lures at machine speed.
Our research all points to the same conclusion: identity remains the main battleground, especially the identities you don’t see. The twist for 2026 is that attackers will increasingly go around people, stealing API keys and tokens from CI/CD pipelines, abusing misconfigured workload identities in cloud platforms, or hijacking AI agents that were granted far more permissions than any human user.
That’s why we’re investing so much in work on ITDR, secrets management, non‑human identities and API security, because that’s where a lot of the “invisible” attack paths now live. In our report on Attack Surface Management, we describe how cloud, SaaS, remote work and DevOps have turned the attack surface into a constantly moving target – across external assets, third parties, and human behavior. Combine that with the explosive data sprawl, and you get a simple picture for 2026: more exposed services, more shadow data, more opportunities for double extortion and “quiet” data theft without any encryption at all.
Where the real surprises are likely to come from
So, if that’s the “same, but worse” part, where should we expect actual surprises?
1. The AI and identity supply chain
A lot of organizations are racing to bolt GenAI into business processes via connectors, plugins, and SaaS integrations. Very few treat that as a new supply chain with its own threat model.
In 2026, we might start seeing incidents where a poorly-designed AI assistant exfiltrates data from multiple systems at once, a prompt injection forces an autonomous agent to abuse its own credentials, or a third‑party model or training pipeline becomes the entry point for a broader breach.
From our perspective, this is just identity and data security in a new wrapper, which is exactly why we keep mapping AI back to existing disciplines like DSPM, ITDR and API security instead of inventing yet another silo.
2. Non‑human identities at industrial scale
The whole reason we’re dedicating so much coverage to Non‑Human Identities is that NHIs already outnumber and often outrank humans in many environments. The surprise for many boards in 2026 will be that a “classic” breach – lateral movement, privilege escalation, data exfiltration – can now happen almost entirely through machines, without any human phishing at all.
Examples of such attack vectors can be as trivial as a forgotten script with domain administrator’s credentials or as complex as a rogue AI agent integrated with ticketing, identity and knowledge systems, quietly doing the attacker’s bidding. When those stories hit the headlines, people will call them “AI attacks”. Under the hood, they are just identity governance failures.
3. Cyber‑physical spill‑over and digital sovereignty
We’ve also been looking at the geopolitical side of all this. Our blog on The EuroStack Vision is a good example: it discusses Europe’s heavy dependence on a small number of non‑European cloud and platform providers and what that means for digital sovereignty.
At the same time, national security agencies are reporting sharp increases in “nationally significant” cyber incidents and are openly debating restrictions on ransom payments and stricter obligations for critical infrastructure.
The 2026 surprise here is not that critical infrastructure or public services can be hit – we’ve known that for years – but how quickly cloud outages, supply‑chain incidents, or targeted campaignscan cascade across countries and sectors when everyone is sitting on the same few providers and the same fragile software stacks.
2026: same, worse, or new?
2026 will bring familiar threats amplified by AI automation, exploding nonhuman identities, and fragile, highly concentrated infrastructure. Ransomware, data sprawl, and identity abuse will intensify, while AI supply chain gaps, machine-driven breaches, and geopolitical tensions will turn old attack patterns into systemic business risks.
From a defensive point of view, none of this invalidates what we’ve been recommending in 2025:
- Treat identity – human and non‑human – as your primary security perimeter
- Maintain continuous visibility into what you actually expose and where your sensitive data really lives
- Invest in resilience: assume compromise, design for rapid recovery, and rehearse what you’ll do when a key SaaS platform or cloud region goes down
If organizations do that, 2026 will still be challenging – but they’ll be dealing with trends they already understand, rather than discovering them for the first time in a crisis.
If you want to dig into these topics with people who are living them every day, I’d strongly suggest joining us at the European Identity and Cloud Conference 2026, taking place May 19–22, 2026 in Berlin, Germany. EIC is our annual gathering point for everyone working on digital identity, security, privacy, and governance in an AI‑driven world. You’ll hear from practitioners, vendors, regulators, standards experts and KuppingerCole analysts – and you’ll have plenty of opportunities to challenge, refine and stress‑test your own plans for 2026 and beyond.