Malware That Rewrites Itself Every Hour: AI-Powered Polymorphic Threats Break Signature-Based Defense

AI-powered polymorphic malware is malicious code that uses artificial intelligence to automatically rewrite itself — often every hour or on every execution — so that no two copies share the same signature, hash, or structure. Because traditional antivirus and signature-based detection depend on recognizing a known pattern, malware that never holds still is effectively invisible to them until after it has already run.

AI-Powered Polymorphic Malware, Explained

This is no longer theoretical. In late 2025, Google's threat researchers disclosed PROMPTFLUX, an experimental dropper that queries the Gemini API roughly once an hour to regenerate fresh, obfuscated versions of itself. Researchers have since observed samples that produced more than 70 distinct variants of themselves in under four hours — analyzing their own structure and rewriting their own code autonomously. The barrier to building a shape-shifting threat has collapsed.

For defenders, the implication is structural, not incremental. You cannot out-detect a threat that mutates faster than you can write detections for it. The only strategy that survives is prevention-first cyber defense — stopping the threat deterministically at the moment of execution, whether or not anyone has ever seen it before. That is the foundation of Morphisec's Automated Moving Target Defense (AMTD).

What Is Self-Rewriting Malware, and How Does It Work?

Self-rewriting malware is malware that regenerates its own code while it runs, producing a brand-new variant each time so that static fingerprints never match. Where classic polymorphic malware simply re-encrypts a static payload around a fixed core, AI-driven strains go further: they use large language models to generate entirely new code blocks, pushing toward true metamorphic behavior with no stable core to fingerprint at all.

The mechanics are now well documented. 

PROMPTFLUX calls an LLM hourly to rebuild itself. A companion sample, PROMPTSTEAL, uses a model to generate one-line Windows commands on demand to harvest documents. Named families such as PromptLock and BlackMamba use LLMs to rewrite their payloads in real time. The common thread: the malware treats an AI model as a code factory, manufacturing novelty faster than any human analyst — or signature feed — can keep up.

• No stable core: Mutation is metamorphic, not just encrypted — there is no constant payload to fingerprint.
• Machine speed: The malware operates autonomously, learning from each failed attempt without human guidance.
• Infinite disguises: Each variant is functionally identical but structurally unique, defeating hash and signature matching.

Why AI Polymorphic Malware Breaks Signature-Based Detection

Signature-based detection works by comparing files and behaviors against a catalog of known-bad patterns. That model has one unavoidable assumption: the threat holds still long enough to be recognized. AI-powered polymorphic malware violates that assumption on purpose.

When code is different on every execution, there is no signature to match, no hash to block, and no consistent behavioral pattern to learn. Even behavioral AI detection — the supposed answer to evasion — is forced into an arms race it cannot win, because the attacker only has to be different, not better. The defender must correctly classify infinite novel variants; the attacker only has to produce one the model has not seen. That asymmetry favors the attacker every time.

This is the same evasion logic behind fileless and in-memory malware, which Morphisec has tracked for years: when there is nothing on disk to scan and nothing familiar to match, scanners and reputation engines run out of road. AI mutation simply industrializes that evasion.

Why Detection-First Security Is Always a Beat Behind

Detection-and-response is reactive by design: it must first observe a threat, then recognize it, then respond. Against malware that rewrites itself at machine speed, every step in that chain arrives late. By the time an alert fires, the variant that triggered it has already been replaced — and the payload has already executed.

Time-to-exploit has accelerated to machine speed, and reactive detect-and-respond simply cannot run that fast. This is not a failure of any single product or team; it is a structural limit of a model built on recognition. The legacy security model assumes the defender can keep a current picture of what 'bad' looks like. AI-powered polymorphic malware makes that picture obsolete the moment it is drawn.

Prevention-First Defense: Stopping Malware Before It Executes

Prevention-first cyber defense flips the model: instead of trying to recognize the threat, it removes the conditions the threat needs to execute. Morphisec's Automated Moving Target Defense (AMTD) morphs the runtime memory environment at process launch, relocating the resources legitimate applications expect and leaving a decoy where the real targets used to be.

A self-rewriting payload can mutate its code an infinite number of times, but it still has to resolve to a real target in memory to run. Under AMTD, that target is not where the malware expects it. 

The attack lands on the decoy and fails — deterministically, at execution, with no signature, no model training, and no prior knowledge of the variant required. Morphisec does not need to know what the malware is. It only needs the malware to try to execute.

• Certainty: Deterministic, not probabilistic — the outcome does not depend on recognizing the threat.
• Pre-execution: Stops the payload at execution, before encryption, exfiltration, or lateral movement begins.
• Low overhead: Lightweight and update-light — no constant signature feeds or heavy scanning required.

This is why AMTD is the core of Morphisec's ransomware protection: it closes the runtime gap that machine-speed mutation pries open, and it does so without competing in the detection arms race at all.

How AMTD Fortifies Your Existing Security Stack

Prevention-first does not mean rip-and-replace. 

AMTD is complementary to the EDR, NGAV, and scanning tools already in place. Those tools remain valuable for visibility, investigation, and known-threat coverage. What AMTD adds is a deterministic backstop for exactly the threats detection cannot see: zero-days, fileless attacks, and now AI-generated polymorphic malware that has no recognizable signature.

In practice, deterministic prevention fortifies detection and response: it stops the unknown payload at execution, then feeds high-fidelity, low-noise forensic data back to the Morphisec platform and the broader stack. The result is fewer alerts, less analyst fatigue, and — critically — protection against threats that have not been written yet.

Prevention Beats Detection Every Time

AI-powered polymorphic malware is the clearest proof yet that the signature era is over. 

When malware rewrites itself every hour, recognition is a losing math problem — the attacker only has to be different, and AI makes being different effortless. You do not out-detect a shape-shifter. You remove the target it is shooting at.

Morphisec's patented AMTD technology does exactly that: it stops threats deterministically, at execution, before the alert ever fires. Against malware that mutates at machine speed, deterministic prevention is not just the better strategy — it is the only one that holds.

See AMTD stop what detection can't.

Book a live demo and watch Morphisec neutralize a polymorphic, self-rewriting payload at execution — no signature required.

About the author

Brad LaPorte | New York

Chief Marketing Officer

Brad LaPorte is a seasoned cybersecurity expert and former military officer specializing in cybersecurity and military intelligence for the United States military and allied forces. With a distinguished career at Gartner as a top-rated research analyst, Brad was instrumental in establishing key industry categories such as Attack Surface Management (ASM), Extended Detection & Response (XDR), Digital Risk Protection (DRP), and the foundational elements of Continuous Threat Exposure Management (CTEM). His forward-thinking approach led to the inception of Secureworks’ MDR service and the EDR product Red Cloak—industry firsts. At IBM, he spearheaded the creation of the Endpoint Security Portfolio, as well as MDR, Vulnerability Management, Threat Intelligence, and Managed SIEM offerings, further solidifying his reputation as a visionary in cybersecurity solutions years ahead of its time. He is based in Morphisec’s New York office at 122 Grand St, New York, NY.