AI is making parts of cyberattacks faster and easier to adapt, but it has not replaced conventional cybercrime. Google Threat Intelligence Group (GTIG) reports that attackers have used AI to research targets, create phishing lures, assist malware development and explore vulnerabilities. Microsoft’s 2025 threat report, meanwhile, emphasizes familiar weaknesses: known security gaps and password attacks. The practical takeaway is to treat AI as an added capability in an attacker’s toolkit—not proof that every attack is novel, autonomous or successful.
How are hackers using AI?
“AI-driven cyberattack” can describe several different things. It might mean an attacker uses an AI tool for one task, such as drafting a convincing message; it might mean AI-generated code or content is incorporated into an attack; or it might mean malware queries a model while it is running. Those are distinct capabilities, and evidence for one does not establish the others.
As an Amazon Associate I earn from qualifying purchases.
GTIG’s reporting gives examples across the attack lifecycle. Its November 5, 2025 threat-tracker report describes actors using AI for reconnaissance, phishing, malware development, command-and-control development and data exfiltration. GTIG also reported an illicit market for AI tooling and social-engineering-style prompts intended to bypass model safeguards. These are the group’s observations, not a count of all attacks or attackers.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors| Use of AI | What it means | What the reporting establishes |
|---|---|---|
| Research and reconnaissance | Using AI to gather or organize information about targets and environments. | Google reported observing AI used for information gathering; GTIG’s November 2025 report described use across reconnaissance and other operational tasks. |
| Phishing and social engineering | Generating or refining lures that are tailored to a target or situation. | Google reported realistic phishing scams, and Microsoft described AI-automated phishing. Neither finding means every phishing message is AI-written. |
| Malware development | Using AI to assist in writing, modifying or adapting malicious code. | GTIG reported malware-development assistance and experimental malware that could use a model during execution. |
| Vulnerability exploitation | Using AI to investigate or help develop ways to exploit software flaws. | In May 2026, GTIG said it believed a threat actor’s zero-day exploit had been developed with AI. That is an attributed assessment, not a confirmed finding that AI independently created the exploit. |
Sources: GTIG, “GTIG AI Threat Tracker: Advances in Threat Actor Usage of AI Tools” (November 5, 2025); Google, “Our new report details the latest ways threat actors are misusing AI” (February 12, 2026); Microsoft, “Microsoft Digital Defense Report 2025”; and GTIG, “GTIG AI Threat Tracker: Adversaries Leverage AI for Vulnerability Exploitation, Augmented Operations, and Initial Access” (May 12, 2026).
#1 Best Overall
Can AI create malware that changes while it runs?
GTIG’s November 2025 report says it identified malware families using large language models during execution to dynamically generate or alter malicious behavior. It named PROMPTFLUX and PROMPTSTEAL among its examples and characterized these capabilities as nascent. This is more specific than using an AI assistant to write code: the malware’s behavior can be influenced by a model while the program is operating.
GTIG’s May 2026 report also described autonomous behavior in PROMPTSPY and AI-supported malware and defense evasion. Those observations show experimentation and evolving capability; they do not establish that malware is broadly operating without human direction or that it can defeat every defense. The distinction matters: an experimental capability reported in particular cases is not the same as a routine feature of cybercrime.
Are AI cyberattacks worse than regular cyberattacks?
AI can raise the speed, volume or adaptability of particular tasks, but the available reporting does not show that AI is behind most cyberattacks. A convincing AI-assisted lure can still depend on a victim responding, and AI-assisted exploit development does not mean that an exploit works against every target. Nor does the presence of AI make an attack autonomous from end to end.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Microsoft’s 2025 report supplies important context: it says many threats targeted known security gaps, including web assets and remote services. It also reports that 97% of identity attacks in Microsoft’s observed data were password-spray attacks. That figure describes Microsoft’s dataset, not all identity attacks globally, and it is not a statistic about AI attacks.
Rank #3
The evidence supports a narrower, more useful conclusion: AI can make some steps in an established attack workflow more efficient or adaptable, while exposed systems and weak account defenses remain relevant. The reporting does not provide a verified global count or prevalence rate for AI-driven cyberattacks, so claims that a particular share of all attacks use AI are not established by these sources.
What do the reports say about AI-enabled attacks on AI systems?
Google’s February 12, 2026 report said GTIG had not observed direct advanced persistent threat (APT) attacks on frontier models or generative AI products. It separately reported model-extraction attacks against private-sector entities. Model extraction is an attempt to obtain information about or reproduce a model; it should not be presented as equivalent to an ordinary intrusion into a company network.
In May 2026, GTIG also reported software supply-chain targeting involving AI environments and dependencies. That is a separate risk from attacking a frontier model directly: software components and surrounding environments can be targets even when a direct attack on the model itself has not been observed. GTIG’s May assessment drew on Mandiant incident-response engagements, Gemini and its own threat-intelligence work, so its reported cases should be understood as the organization’s findings rather than a census of the threat landscape.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What is the reported AI-assisted zero-day—and how certain is it?
In its May 12, 2026 report, GTIG said it identified a threat actor using a zero-day exploit that GTIG believed had been developed with AI. A zero-day is an exploit for a vulnerability that was not previously known to the software vendor or for which a fix was not yet available at the relevant time. The wording is important: GTIG described its belief about how the exploit was developed; it did not establish that AI alone created it.
Best Value
GTIG also said the actor planned mass exploitation and that its proactive discovery may have prevented that plan. The report therefore documents a serious suspected use of AI in exploit development, alongside an intervention—not proof that a mass campaign succeeded or that AI has made zero-day attacks commonplace.
How can you protect yourself from AI-assisted phishing and account attacks?
Use stronger sign-in protection for important accounts. CISA recommends multifactor authentication (MFA) and favors phishing-resistant methods. A physical security key is one option where the account and devices support it. Check compatibility before relying on a key, and set up the account’s recovery options so losing the key does not lock you out. An authenticator app is another MFA option described in CISA guidance; availability and protection depend on the service and setup.
- Check whether MFA is available. In the account’s security or sign-in settings, enable MFA and review the methods the service supports.
- Prefer a phishing-resistant method when supported. Consider a compatible physical security key, and verify that it works with the account and the devices you use to sign in.
- Prepare recovery before you need it. Review the account’s recovery process and keep an appropriate backup method available, especially if the key is your primary sign-in factor.
- Continue to assess messages and requests carefully. AI-generated wording can be polished, so good grammar or a familiar tone is not proof that a message is legitimate. Verify unexpected requests through a separate, trusted channel rather than relying only on the message.
CISA’s MFA guidance treats security keys as an authentication defense, not a universal shield. A key can help protect sign-in to compatible services; it does not stop malware, software exploitation or every form of social engineering.
Free tools Windows power users keep installed
One-click scans. No signup required.
Quick Recap
What should you take away from the AI threat reports?
- GTIG, Google and Microsoft have reported real uses of AI in cyber operations, including research, phishing, malware work and other attack tasks.
- Some model-assisted malware and AI-related exploit activity are emerging capabilities, not evidence that attacks are universally autonomous.
- Microsoft’s reporting shows why conventional defenses still matter: known security gaps and password attacks remain prominent in its observations.
- For personal account protection, enable MFA and choose a phishing-resistant option where your service supports it; keep recovery and compatibility in mind.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
