Cybersecurity trends 2026: agentic AI, leakware, and how to defend

Every year the attack picture shifts a little — and 2026 has a few moves worth preparing for rather than chasing after. The biggest change is not a new strain of malware but the way attackers organize their work: faster, more automated, and aimed straight at the things that matter most, namely data and identity. This piece walks through the trends worth watching in 2026 and how a Vietnamese organization can arrange a layered defense — without panic, but without complacency either.

Agentic AI: attacks get automated

A campaign used to take many hours of human effort: scanning service ports, testing for vulnerabilities, writing exploit scripts, drafting phishing emails. With agentic AI, most of those repetitive steps are handed to software that runs on its own. An attacker can stand up an automated pipeline that scans broadly, sorts which targets are easiest to compromise, then tries a sequence of exploit variants with almost no extra human labor.

For defenders this has two practical consequences. The first is speed: the window between a vulnerability being disclosed and it being scanned and exploited at scale keeps shrinking, so patching a few weeks late is far riskier than it used to be. The second is scale: small targets that were previously "not worth an attacker's time" stay in the crosshairs because the cost per target drops close to zero.

It is worth remembering that AI is no magic wand for the attacker either. It amplifies techniques that already exist and rarely invents an entirely new class of attack. So the foundation of defense is still doing the basics well: shrink the attack surface, patch quickly, and automate monitoring so you can keep pace with the other side.

Leakware: from encryption to data theft

Classic ransomware works by encryption: lock the victim's data, then demand a ransom to unlock it. The counter has been clear for a long time — with good backups that are isolated and tested for recovery, most encryption attacks lose their sting, and the business simply restores from backup.

Leakware, or data-extortion, inverts that logic. Instead of (or alongside) encrypting, the attacker exfiltrates the data and threatens to publish it unless paid. Backups are no longer the rescue, because the problem is not whether you can restore your data but that sensitive data is now in someone else's hands. The risk shifts from "operational disruption" to "information leak," dragging along legal consequences under personal-data protection rules and lost trust with customers and partners.

This shift changes where defense should focus. Backups are still needed but no longer sufficient. The central questions become: where does sensitive data live, who can reach it, and how do you detect when an unusually large volume of data is pulled out of the system. This is exactly where keeping data under local control and reducing exposure earns its keep.

Identity is the front door

As systems get better protected at the network layer, attackers pivot to the path of least resistance: logging in with valid credentials rather than breaking down the door. A leaked username and password, a hijacked session, or an employee tricked into approving an authentication prompt — each drops the attacker inside as a real user, hard to distinguish from normal activity.

Notably, even multi-factor authentication (MFA) is no longer an absolute shield. Techniques such as bombarding a user with prompts until they approve one by mistake, or a phishing page sitting in the middle to steal an authenticated session, show that MFA must be deployed correctly rather than just switched on. Identity management in 2026 therefore has to go beyond passwords: least privilege so a single compromised account does not open the whole system, and behavioral monitoring to catch anomalies such as logins from unfamiliar locations or access to data outside someone's job scope.

Supply chain and SaaS: third-party risk

Most modern systems are assembled from pieces the organization did not write itself: open-source libraries, SaaS services, vendor accounts, integrations wired up through APIs. Each piece is a door, and any door is only as strong as the weakest link in the chain. A compromised dependency, a breached supplier, or a leaked integration token can all become an entry point your firewall never sees.

SaaS risk has an easily overlooked angle too: the organization's data ends up scattered across many external platforms, each with its own permission settings. A folder accidentally shared publicly, an old account never revoked, a default configuration that is too permissive — these are gaps that need no advanced skill to exploit. Taking inventory of where important data is processed, and keeping the most sensitive of it under your own control, is the most compact way to cut risk. It is also why many organizations lean toward self-hosting their core systems: fewer third-party dependencies means fewer doors to guard.

APIs and bots: an expanding attack surface

Applications today talk mostly over APIs, and APIs are attractive targets because they speak the language of data directly. The common problems are rarely exotic vulnerabilities but basics left undone: an endpoint exposing more data than it needs, a missing authorization check that lets one user read another's data, or an API with no rate limiting that lets an attacker probe at leisure.

Running alongside that is automated bot traffic. Not every bot is malicious, but malicious ones do things that are expensive for the victim: stuffing leaked credentials at scale, scraping data, abusing functionality, or simply overwhelming the system. Separating real traffic from bots, and blocking abuse at the layer in front of the application, lightens the load on everything behind it and cuts out most of the noise from automated attacks.

This is precisely where a web-and-API protection layer (WAAP) earns its place: sitting in front of the application to filter malicious requests, absorb DDoS, manage bots, and protect API endpoints. Manta's Orca WAAP is a self-hosted option along these lines — you hold the data and the filtering rules yourself rather than routing all your traffic through an external service. For a deeper look at the WAAP model, see the post on WAAP security for businesses.

Layered defense: no silver bullet

What all these trends share is that no single product or measure solves everything. The durable approach is still layered defense: assume every layer can be bypassed, and stack enough of them that one breach does not collapse the whole system. For organizations handling sensitive data or in the public sector, keeping core systems self-hosted also narrows external dependencies and keeps data under control — a meaningful edge against both leakware and supply-chain risk. This theme is covered in more depth in the post on enterprise cyber defense for Vietnamese businesses.

2026 defense checklist

• Web and API surface. Put a WAAP layer in front of the application, enable DDoS protection and bot management, and review API endpoints for excess data exposure or missing authorization checks.
• Fast patching. Shorten the time from disclosure to patch; prioritize components exposed to the Internet.
• Identity. Deploy strong authentication correctly, enforce least privilege, revoke stale accounts, and monitor for anomalous logins.
• Data. Inventory where sensitive data lives, encrypt it at rest and in transit, reduce exposure, and alert on unusually large outbound data flows.
• Supply chain. Review dependencies and vendor permissions; limit the scope of integration tokens; prefer keeping core systems under your own control.
• Incident response. Keep a clear process and rehearse it; maintain isolated backups and test recovery regularly; prepare a playbook for the case where data is stolen, not just encrypted.
• Continuous monitoring. Centralize logs and set alerts so you can keep pace with automated attacks — early detection matters more than tall walls.

None of these trends demands tearing everything down and starting over. Most of it is doing the basics well and re-ordering priorities to match how attackers actually operate in 2026. If your organization needs to assess its current posture and find where the gaps still are, book a consultation and we will review it together.