ATT&CK v20 Maps Weaponized AI

Did you know attackers are already using your own enterprise AI against you?
MITRE has just updated its ATT&CK framework to version 20.
The ATT&CK framework is the global reference map used by defenders and analysts to understand how attackers operate. Each version reflects what is actually happening.
This time the message is clear: artificial intelligence is no longer just a defensive tool. Adversaries have turned it into a weapon, and MITRE is documenting it.

Prompt injection now has its own dedicated space on the map.
ATT&CK v20 introduces two distinct techniques for prompt injection. The first covers malicious instructions entered directly into an LLM interface. The second, more covert, describes instructions hidden inside documents that the model ingests without review. Additionally, a new technique documents how attackers can leverage an organization’s own AI agents to steal internal data.
They are also poisoning the web content that assistants query.
Two sub-techniques complete the AI block: one covers web content poisoning to manipulate search engine and intelligent assistant recommendations, and another describes how attackers interfere with LLM-assisted malware analysis. In parallel, ATT&CK v20 adds a technique for stealing application passwords that allow login without going through MFA.
Adversaries are already bypassing AI, MFA, and verification bots.
The ICS domain also receives critical updates. A new campaign documents attacks against PLCs at more than thirty water facilities in Minnesota, leaving operators locked out. Attribution remains unconfirmed. A second campaign records an AI-assisted intrusion against Mexican government institutions that reached a municipal water plant.

Threat actors also have new entries.
ATT&CK v20 adds Nimbus Manticore, an espionage group linked to the Iranian IRGC, and updates the entries for Salt Typhoon and APT38. The Gentlemen make their debut as ransomware-as-a-service operators. Defenders receive fifteen new detection strategies backed by thirty-eight analytics.
AI as an attack vector is now official.
MITRE cataloguing these techniques in ATT&CK is not an academic exercise. It means security teams around the world will use them as a reference for audits and red team exercises. When something enters ATT&CK, it moves from being an emerging threat to a documented risk that organizations must address in their threat models. The abuse of AI agents has just crossed that line.
The mobile domain is not left out of the update either.
In the Mobile domain, MITRE reinstates a technique for distributing malicious apps outside official stores, absent since version 11. It also adds NGate, a malware that intercepts payment card data via the phone’s NFC and enables cash withdrawals at ATMs without the physical card.

ICS analytics are designed to filter out routine maintenance.
That detail matters greatly: the new analytics for industrial environments filter out the usual noise from engineering and maintenance activities, reducing false alarms and making real alerts harder for operations teams to ignore.
ATT&CK v20 turns AI into a battlefield.
Until now, security frameworks treated AI primarily as a defensive tool. With this version, MITRE formally acknowledges that LLMs, AI agents, and intelligent assistants are also an attack surface. Any organization that has deployed AI tools without reviewing its threat models has a gap to fill.
The map has changed, and attackers already know it.
The question now is whether defenders will update their controls before someone exploits these newly catalogued techniques in their environment.
What you can do
- ✓Review whether your AI agents have access to sensitive data.
- ✓Verify that MFA does not rely on application passwords.
- ✓Incorporate ATT&CK v20 techniques into your threat model.
Is your team prepared to also defend the AI systems you have already deployed?
Security is not improvised, it is audited. At Nacata Security we detect vulnerabilities and protect your company, because a single flaw can cost you everything you have built.
Related articles
Nacata Security, reach out to us anytime
We are Nacata Security, get to know us
web: nacata.io
email: info@nacata.io
Phone: 919930793
LinkedIn: Nacata Security
The intersection of artificial intelligence and cybersecurity: attacks that use AI to deceive or impersonate, security of models and conversational assistants, and new AI-based threat detection tools.
RATING
7.1
Who are we?
At Nacata Security we are an offensive cybersecurity company specialized in audits and penetration testing.
We detect, assess and help mitigate the vulnerabilities of your systems, networks and applications before a real attacker exploits them, offering 360º defense tailored to each client.
We’d be glad to get in touch with you for whatever you need.




