Skip to Content

A Year in Review: 2026 Cyber Threats

In 2026, three of our four predictions from last year's held. Here’s a look back at how each one evolved.

2026 Prediction Result and Grade

  1. Prediction 01
  2. Prediction 02
  3. Prediction 03
  4. Prediction 04
01Threat actors will abuse and commoditize commercial AI pen-testing tools in 2026, following the Cobalt Strike playbook, to increase the speed, precision, and scale of their campaigns.B
Result

Commoditization occurred rapidly, but through a different mechanism than anticipated. The original forecast expected adversaries to follow the Cobalt Strike model by cracking, leaking, and reselling commercial software. In reality, threat actors bypassed that hurdle entirely because developers published autonomous AI pen-testing frameworks openly and free of charge on public repositories. Towards the end of 2026, tools like ARTEX and Cairn were actively weaponized in intrusions against hundreds of organizations before developers attempted to pull the code. While the operational impact materialized as predicted, the open-source delivery path differed fundamentally from the commercial cracking playbook.

02Specialized cybercriminal groups will target open-source maintainers in 2026, using social engineering to compromise the human element and inject malicious code into widely used repositories.A+
Result

This attack pattern escalated sharply throughout the year. Building on activity observed in late 2025, campaigns like the "Shai-Hulud" npm worm targeted repository maintainers through credential phishing to harvest publishing tokens and poison downstream package updates. Because the worm's propagation routine was short and publicly reconstructed, adversaries routinely forked and deployed clones across multiple package registries throughout 2026, proving this technique is now an established supply-chain playbook.

03China-backed advanced persistent threat (APT) groups will exploit Beijing's dominance over the Internet of Things (IoT) supply chain to embed persistent backdoors in Western critical infrastructure, enabling long-term espionage and disruption.C
Result

Chinese state-sponsored pre-positioning across critical networks was widely documented throughout 2026, as seen in the suspected China-linked espionage cluster that ReliaQuest tracks as “OP-512.” However, deliberately manufactured hardware backdoors didn’t materialize. Adversaries instead exploited unpatched software vulnerabilities in edge network devices, VPN gateways, and application delivery controllers, and hijacked residential routers to proxy traffic.

04Emerging cybercriminal groups will embrace loud branding to attract affiliates while established actors prioritize secrecy, creating prioritization problems for security teams.B−
Result

The forecast anticipated a sharp division between loud new entrants and secretive legacy actors. In practice, emerging groups embraced aggressive marketing by publishing compromises in massive bulk dumps, frequently listing upwards of 40 victims in a single day to manufacture industry notoriety and recruit affiliates. At the same time, the wider ecosystem adapted through formal operational consolidation. Rather than competing as isolated brands, major players formed federated alliances, exemplified by access broker TeamPCP feeding high-volume credentials directly into the ransomware infrastructure run by Vect.[^1] The market ultimately balanced aggressive public posturing from newcomers with resilient, collaborative cartel partnerships engineered to withstand international law enforcement takedowns.[1]

  1. The Security Supply Chain Turned on Itself

    Security and DevOps products hold privileged positions inside continuous integration (CI) pipelines with broad credential access, making them an exceptionally efficient route into software supply chains. In 2026, a threat group tracked as TeamPCP ran a three-stage supply-chain compromise demonstrating this exposure. The adversary breached the vulnerability scanner Trivy, used those privileges to compromise the infrastructure-as-code scanner Checkmarx KICS, and poisoned the LiteLLM model-routing proxy. By injecting a startup hook that harvested environment variables and cloud keys, the campaign exfiltrated roughly 153GB of credentials across approximately 2,488 corporate domains.

    TeamPCP's three-link chain through the security industry's own tooling[2]
    1. Link 1Trivy
    2. Link 2Checkmarx KICS
    3. Link 3LiteLLM
    ~153GB
    of credentials exfiltrated
    ~2,488
    corporate domains
  2. Vishing and Device-Code Phishing Outflanked MFA

    Through 2026, the most effective intrusions against enterprise identity bypassed MFA through direct voice interaction rather than technical exploits. To evade corporate defenses, threat actors targeted employees on their personal devices. Adversaries mapped organizational rosters and harvested personal phone numbers through commercial data aggregators, ensuring the initial social engineering call happened off the corporate network, where security tools couldn't monitor it. Both vishing and device-code phishing succeed because the user completes the authentication workflow themselves, producing an entirely legitimate session token. "ShinyHunters" scaled this by branding generic domains at the subdomain level, calling employees on their personal phones under a help-desk pretext, and prompting them to approve mobile MFA prompts while proxying single sign-on (SSO) sessions. Meanwhile, "Helix" paired phone pretexts with device-code phishing, tricking users into entering an attacker-supplied pairing code into a legitimate Microsoft sign-in page to bypass conditional access and automate SharePoint exfiltration. Once inside, these groups immediately registered secondary authentication factors, ensuring that subsequent password resets or session revocations left their access intact.

  3. Hacktivist Attacks Turn Destructive

    Hacktivist groups increasingly crossed the boundary from digital espionage into physical disruption, intentionally targeting commercial and operational technology (OT) systems to exert economic and public pressure. In late 2026, an Islamic Revolutionary Guard Corps (IRGC) cyber unit targeted internet-connected commercial refrigeration controls across eight European nations, spoiling perishable food across supermarkets and cold-storage facilities.[3] Similar pressure tactics hit another sector. As detailed in ReliaQuest reporting on the business impact of expanding geopolitical cyber threats, the pro-Iranian group “Handala” claimed the destruction of over 200,000 devices at a medical technology company, showing that private-sector organizations don't need to be direct participants in a geopolitical conflict to suffer operational fallout.

The biggest lesson of 2026 was how fast experimental techniques became routine tradecraft.

All four of our predictions for 2027 share an economic thread.

For threat actors, the rise of open-source models has made capable AI cheap, accessible, and easy to host locally without vendor safety guardrails, sharply lowering the technical bar for what an attacker needs to bring to an intrusion. For enterprises, supporting that same capability at scale has become increasingly costly, forcing compromises that introduce new governance risks. In this report, we use the developments that defined 2026 as the foundation for what enterprises are likely to face in 2027. While these forecasts are necessarily speculative, each prediction is grounded in our own incident data, dark-web collection, and published reporting, to give a realistic outlook on the year ahead.

2027 Predictions Threat Report

Short on Time? Read It Later.

Get the full report as a PDF in your inbox and read it when it suits you.