AI Cyber Attack Statistics 2025

Over the last few years, artificial intelligence has gone from a technology of the future to a mainstream tool that anyone can use. AI is now widely available and has helped many businesses work more efficiently.

However, advancements in AI have also introduced dangerous new cybersecurity risks that every business should know about. These risks come from new vulnerabilities in AI systems and the growing use of AI-driven methods by attackers.

We've rounded up the latest AI cyberattack statistics so you can understand the scope of these new threats. Every number on this page links to its primary source — if we couldn't verify it, we left it out.

The growing role of AI in cybercrime

AI isn't just a tool for defense — it's increasingly being used to power more sophisticated cybercrime. Attackers now use advanced models and machine-learning techniques to evade traditional security controls and personalize attacks at scale.

This shift means organizations must rethink how they assess vulnerabilities and strengthen incident response. Reactive approaches aren't enough when attacks can adapt and evolve in real time.

AI phishing attack statistics

Phishing is one of the most common ways hackers exploit AI. Attackers now use AI tools like ChatGPT to mimic writing styles and avoid detection, taking advantage of vulnerabilities in email filters and user habits. The old advice of “look for typos” is dead.

The numbers explain why even savvy users get caught:

And the time-to-craft a convincing phishing email has collapsed. Generative AI tools help attackers compose phishing emails up to 40% faster, which means the same attacker can run more campaigns against more targets. 65% of phishing attacks now target organizations rather than individuals.

AI deepfake statistics

A deepfake is a digitally generated image or video made to look and sound real. Large language models and generative video have collapsed the cost of producing one, and the financial industry has become the most-attacked target.

• Only 0.1% of people can consistently identify a deepfake, even when primed to look for one (iProov tested 2,000 UK and US consumers in 2025).
• 53% of financial professionals have experienced an attempted deepfake scam (Regula Forensics survey, 2024).
• Per the Signicat report, deepfakes are now the most common form of digital identity fraud in financial services across Europe, accounting for roughly 6.5% of all fraud attempts — up from 0.1% three years ago.

And it's not theoretical — there's a concrete case that should be on every CFO's mind.

AI password-hacking statistics

AI has changed the economics of brute-forcing passwords. What used to take days now takes minutes for the average reused password.

The reason this works: 94% of leaked passwords are reused or duplicated across multiple sites, per a 2025 Cybernews study of 19 billion exposed credentials. Crack one, and you've cracked the user's entire digital identity.

AI voice-cloning statistics

Voice cloning takes a short recording of someone's voice — typically pulled from social media, podcasts, or YouTube — and uses it to generate convincing false recordings of that same voice. It's the engine behind a growing wave of phone scams.

The McAfee “Beware the Artificial Imposter” report surveyed 7,054 adults across seven countries and found that a quarter of adults have personally experienced or know someone who experienced an AI voice scam — and 77% of victims lost money. Of those who lost money, more than a third lost between $500 and $3,000; 7% lost between $5,000 and $15,000.

And per a peer-reviewed study published in Nature Scientific Reports, participants correctly identified a voice as AI-generated only about 60% of the time — and matched the perceived identity of an AI-generated voice to its real counterpart 80% of the time.

AI in cybersecurity defense

AI isn't just an attack tool. The cybersecurity industry is investing heavily in AI for threat detection, response automation, and vulnerability discovery.

Per TakePoint Research, 80% of industrial cybersecurity professionals believe the benefits of AI in security outweigh the risks, and companies using AI-driven detection report identifying threats up to 60% faster than with traditional methods.

Translation: your firm is going to depend on vendors and partners for AI-powered defense, whether you plan for it or not. The right time to ask what your MSP is using and where the gaps are is before the next attack — not after.

What to do about it

The stats above all share a pattern: AI has lowered the cost and raised the quality of every attack type that targets human judgment. Phishing emails that used to be obvious are now polished. Voices that used to be unreliable are now convincing. Faces that used to require Hollywood budgets are now generated for free.

Three practical defenses for any business, especially CPA firms and other regulated industries:

1. Process, not vigilance. Don't rely on humans recognizing an AI-generated message or call. Build approval processes that don't bend based on who appears to be asking — wire authorizations require a callback to a known number, not the number that called you.
2. MFA everywhere. Especially for email, accounting software, and any system that touches money or client data. A password that took 30 seconds to crack still won't work without the second factor.
3. AI-aware security training. Annual training was enough when phishing emails had typos. Now it needs to be quarterly, and it needs to include voice-cloning and deepfake examples — not just email screenshots.

Whether you're running a CPA firm or any other regulated business, the threat model has changed. The good news: most of the defense is operational, not technical.