Loading prices…
🩸BEARISH

Google Finds Autonomous AI Exploit Tool Used by Russians

Autonomous exploitation raises systemic risk across banks and crypto, while North Korean-linked intrusions have helped drain an estimated $12.8B.

Google Finds Autonomous AI Exploit Tool Used by Russians
Google Finds Autonomous AI Exploit Tool Used by Russians
Google Finds Autonomous AI Exploit Tool Used by Russians
Google Finds Autonomous AI Exploit Tool Used by Russians

Google’s Threat Intelligence Group said it found “exploit_agent,” an AI hacking tool that discovered and exploited software vulnerabilities without human intervention. The tool monitored popular software for patches, compared code before and after each update, wrote an exploit and ran it automatically. Google called it the first known case of attackers fully automating vulnerability discovery and exploitation without a human in the loop.

Why it matters

The tool connected to leading commercial and local LLMs, allowing attackers to automate a process that previously required specialist human effort. It could also brute-force passwords, map internal networks, install webshells and retain access through hidden control panels. Russian-language comments on the exposed server referred to “LLM API keys for auto-exploitation,” highlighting the risk of poorly protected credentials enabling autonomous attacks.

The wider disclosure trend is also worsening. Google said overall disclosures and exploits nearly doubled between January and August 2026. Zero-day exploits averaged eight per month in 2025 and 11 per month in 2026 before rising to 22 in August. Autonomous tools could shorten the interval between a patch’s release and its weaponization.

Market impact

The direct threat to crypto firms was demonstrated by Google’s attribution of AI-powered intrusions to MIDNIGHT NEPTUNE, a North Korean group it tracks. The attackers posed as recruiters, altered internal code repositories and used a backdoor named SOMBERMEME to target developers opening Claude Code. Google also identified the open-source attack tools CAIRN and PentAGI in real intrusions.

Separately, blockchain forensics firms and the FBI attributed the roughly $6.3B Binance hack to North Korea-linked hackers. TRM Labs estimated that North Korean-linked entities have stolen about $12.8B in cryptocurrency. The convergence of autonomous exploitation, poisoned developer tooling and state-linked cryptocurrency theft raises the risk that attacks become faster, more scalable and harder to contain before patches can be deployed.

Frequently asked questions

  1. What did Google's exploit_agent tool do?

    It monitored software for patches, compared code before and after updates, generated an exploit and executed it without human intervention.

  2. What did the Russian-language comments reveal?

    Translated comments referred to LLM API keys for auto-exploitation and instructed operators to enter the keys directly.

  3. How did MIDNIGHT NEPTUNE target cryptocurrency firms?

    The group posed as recruiters, altered internal code repositories and used the SOMBERMEME backdoor to target developers opening Claude Code.

  4. How quickly did zero-day exploit activity accelerate?

    Monthly averages rose from 8 zero-days in 2025 to 11 in 2026, then jumped to 22 in August 2026.

  5. How much cryptocurrency have North Korea-linked actors stolen?

    TRM Labs estimated about $12.8B. Blockchain forensics firms and the FBI attributed the roughly $6.3B Binance hack to North Korean hackers.

Source attribution
Aggregated from CoinDesk · Verified · Last refreshed 16m ago
Open original →