Skip to content
Featured on Startup Fame

AI finds zero-day flaws for $4 and launches attacks without human hackers

AI models can now find zero-day vulnerabilities in minutes for less than $5, dramatically lowering the bar for attackers. Worse, a newly discovered AI system, exploit_agent, can turn security patches into fully automated exploits—no human hacker needed. These same tools also give defenders a fighting chance, but only if they act faster than the attackers.

Editor, Lazyfounder

Published 5 min read
AI finds zero-day flaws for $4 and launches attacks without human hackers
Image: blackdovfx via Stock / Getty Images Plus via source

30 SEC SUMMARY

  • AI now finds critical software flaws in minutes for as little as $4.24, a fraction of what traditional methods cost.
  • A newly discovered AI system, exploit_agent, automatically turns security patches into working attacks—no human hacker required.
  • Plummeting AI token prices, down 47% each quarter since 2023, have put advanced hacking tools within reach of inexperienced attackers.
  • Attackers increasingly target n-day vulnerabilities—flaws already patched but still unpatched in many systems—because AI can exploit them faster than zero-days.
  • AI-powered misinformation campaigns now manage vast networks of fake social media accounts and generate deceptive images at scale.

KEY HIGHLIGHTS

  • AI models found a zero-day flaw in Google’s V8 engine in 13 minutes for $4.24, the fastest and cheapest discovery on record.
  • Only three of seven AI models tested succeeded, showing uneven but rapid progress in AI-driven vulnerability detection.
  • AI token prices have fallen 47% per quarter since 2023, putting advanced hacking tools in reach of less skilled attackers.
  • The exploit_agent framework autonomously turns security patches into exploits, targeting French websites across media, entertainment, and education.
  • Attackers increasingly focus on n-day vulnerabilities—patched flaws still present in unupdated systems—because AI tools can exploit them faster.
  • AI-powered influence campaigns now manage over 100,000 social media accounts and generate realistic fake images to spread misinformation.

AI uncovers zero-day flaws in minutes for under $5

Xint.io researcher Taisic Yun tested seven AI models released since February 2026 to see how quickly they could spot a zero-day flaw in Google’s V8 engine. The bug, CVE-2026-102321, allows remote code execution—but only three models delivered. OpenAI’s GPT 6 Sol, GPT 5.6 Sol, and Anthropic’s Claude Opus 5 each found the vulnerability, with costs ranging from $4.24 to $43.40.

GPT 6 Sol was fastest. A single prompt and 13 minutes later, it flagged the flaw for $4.24. GPT 5.6 Sol needed two prompts and 44 minutes. Claude Opus 5 identified it on the first try but ran up a $43.40 bill. The four open-source models—DeepSeek and Z.ai—failed to spot the bug at all.

The tests showed that while some expertise is still needed to guide these models, top-tier skills are no longer required. Falling prices are also helping: AI token costs have dropped 47% every quarter since 2023, putting these tools within reach of less skilled attackers.

Exploit_agent turns security patches into automated attacks

Google’s Threat Intelligence Group stumbled upon an exposed staging server running exploit_agent, a framework that turns security patches into fully automated exploits. The system connects to both cloud-based and local large language models, removing any need for human oversight.

Once exploit_agent detects a patch announcement, it compares the old and new code, identifies the likely flaw, and then drafts and executes an exploit—all on its own. The same server also held stolen data from victims, mostly French websites tied to media, entertainment, and education. A Russian-language comment in the configuration file instructed users to enter API keys directly, not through chat.

Attackers favor patched flaws that remain unfixed in many systems

Google’s latest report, Agentic Threat Activity Crosses a Significant Threshold, reveals a shift: attackers now prefer AI-driven exploits of n-day vulnerabilities—flaws already patched but still unpatched in many systems. The reason? The data from a patch is finite, making it easier for AI models to analyze than hunting for unknown zero-days.

Security teams proved how fast these tools work. Within a day of PaperCut’s flaw disclosure, researchers at Huntress Labs used large language models to generate working exploits. GreyNoise Intelligence tracked a campaign that exploited the same flaws just four days later, compromising hundreds of organizations and gaining domain administrator access within minutes.

AI tools appear in live attacks and misinformation campaigns

Open-source AI frameworks are already being used in the wild. A threat actor deployed CAIRN during an intrusion into the U.S. aerospace sector. Another actor used a customized version of PentAGI to scan organizations across Asia for vulnerabilities between late August and October. These frameworks leave noticeable traces, making them easier to detect than traditional attacks.

AI is also being used to scale misinformation. One platform, built using Google’s Gemini API, now runs over 100,000 social media accounts. It generates fake photos with deliberate camera flaws to make them appear authentic. A suspected North Korean group, UNC6836, used AI-generated images—complete with Google’s SynthID watermarks—in fake recruiter profiles targeting cryptocurrency firms.

What this means

Lazyfounder analysis — our interpretation, not reported fact.

AI tools are shifting cybersecurity from a contest of skill to a contest of speed—defenders must patch vulnerabilities faster than attackers can weaponize them, or risk losing the race entirely.

Key takeaways

  • AI slashes the time and cost to find zero-day flaws, handing powerful tools to attackers who lack deep technical skills.
  • Autonomous frameworks like exploit_agent remove the human bottleneck, turning patches into exploits within hours.
  • N-day vulnerabilities—already patched but not yet fixed everywhere—are becoming the favorite target for AI-driven attacks.
  • Organizations that delay updates now face a higher risk of compromise, even from low-skilled attackers using off-the-shelf AI tools.

Sources

  1. ZDNET · 2026-11-16
    AI finds zero-day flaws for just $4 now – even mediocre hackers can cash in
  2. SiliconANGLE · 2026-11-16
    Google says an AI framework turned security patches into exploits with no human in the loop

This story is an original summary drafted with AI by Lazyfounder from the reporting listed above and checked by automated validation. Facts are attributed to their original publishers; sections marked as analysis are Lazyfounder's. Where a source is in another language, facts were machine-translated and quotations are reported, not reproduced. Read the original coverage via the links, and see our AI policy and corrections policy.

About the author

Editor, Lazyfounder

Tarun Mottlia edits LazyFounders, covering Indian startups, funding rounds, AI and product launches. Every story on the site is AI-assisted and checked against its cited sources before publication.

More stories by Tarun Mottlia

Get the LazyFounder Brief

Startup, funding and AI news in a five-minute read. Join the early-access list.

Lazy Founder - Powered by Blogy.in

Contact us

Have a story tip, correction or partnership idea?

Write to us at tarun.kumar@blogy.in or talk to the founder directly. We read every message.

Contact us