<figure class="wp-block-image size-large"><img width="325px" height="183px" srcSet="https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-375x212.webp 375w, https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-662x374.webp 662w, https://d2ly46a6b1u3g6.cloudfront.net/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks.webp 686w" sizes="(max-width: 425px) 150px, 450px" src="https://ll-blog-dev.s3.ap-southeast-2.amazonaws.com/wp-content/uploads/2026/02/09115132/News-Google-Brings-Agentic-AI-to-Chrome-with-Auto-Browse-for-Smarter-Web-Tasks-662x374.webp" alt="" class="wp-image-9246"/></figure>


Google&#x27;s Threat Intelligence Group just spotted something new. Hackers used AI to develop a zero-day exploit. It&#x27;s the first known case of its kind.
The attack hit an open-source system administration tool. Attackers bypassed two-factor authentication after stealing login credentials. <a href="https://cloud.google.com/blog/topics/threat-intelligence/ai-vulnerability-exploitation-initial-access" target="_blank" rel="noopener">Google&#x27;s Threat Intelligence Group</a> confirmed the details.
Google has &quot;high confidence&quot; an AI model helped discover and weaponize the flaw. The evidence? Exploit code contained what looked like an AI hallucination. Those are errors typical of language model outputs. The formatting patterns matched model training data.
The vulnerability wasn&#x27;t a simple coding error. No memory corruption. This was a logic flaw. A developer hardcoded a faulty trust assumption into the system. Advanced large language models can identify these higher-level security weaknesses. That makes AI a powerful tool for attackers hunting vulnerabilities.
Google&#x27;s report didn&#x27;t name the threat actors. But the company noted something important. China and North Korea have shown strong interest in leveraging AI for vulnerability research. Google also flagged specific malware families. PROMPTFLUX, HONESTCUE, and CANFAIL are already using language models for defense evasion. They generate decoy code to hide malicious behavior.
Threat actors are scaling their AI-assisted operations. They&#x27;re using pooled API keys. They&#x27;re cycling premium accounts. They&#x27;re bypassing model guardrails. Google documented all of it.
This marks a significant evolution in cyber threats. AI&#x27;s moving beyond supporting roles. It&#x27;s not just phishing campaigns anymore. Not just social engineering scams. It&#x27;s now doing direct exploit development. That shift raises particular concerns for systems relying on multi-step authentication. Cryptocurrency platforms depend on account security. They&#x27;re in the crosshairs.
Google warned that attackers are targeting the broader AI ecosystem itself. They&#x27;re probing autonomous skills frameworks. They&#x27;re testing third-party connectors that interact with AI models. The good news? Attackers haven&#x27;t breached the core security logic of frontier AI models. Not yet.
For the cryptocurrency sector, this represents a new category of risk. Crypto platforms depend heavily on secure authentication. They need system integrity. AI tools are getting more sophisticated at identifying logic flaws in authentication systems. Platforms may need to reassess their security architectures. Traditional two-factor protections might not be enough.

Japan Cuts Crypto Tax to 20%, Reclassifies Digital Assets

Binance, Kraken, and Bybit Tokenize SpaceX IPO Equity

Humanity Protocol Loses $36M in Cross-Chain Key Compromise

Coinbase and Cardless Launch USDC-Collateralized Credit Card

Google Detects AI-Driven Zero-Day Exploit in Cyber Attack

Google's Threat Intelligence Group discovered hackers using AI to create a zero-day exploit, marking the first confirmed case of AI-assisted cyber attack.