1. 首页
  2. ›Google: AI Agent Achieves First Fully Autonomous Vulnerability Exploitation Without Human Oversight

Google: AI Agent Achieves First Fully Autonomous Vulnerability Exploitation Without Human Oversight

N
nashnova research
·今天发布·阅读约 8 分钟

Google's threat-intelligence team has documented the first case of attackers using an AI multi-agent framework to discover and exploit software vulnerabilities with zero human intervention — zero-day exploits now average 11 per month, up from 8.

01

What exactly happened?

•
Google's Threat Intelligence Group (GTIG) found a multi-agent framework — a set of AI programs dividing tasks automatically — deployed on an exposed test server.
•
The tool compares code before and after a patch, identifies potential vulnerabilities, then generates and runs exploit code, all without step-by-step human commands.
•
This means → the full chain of "find a flaw → write an exploit → launch an attack," once a manual craft, can now run end-to-end on autopilot.
02

How bad are the numbers?

•
Google's vulnerability disclosures from January to August 2026 more than doubled year-on-year.
•
Zero-day exploits — attacks that hit before the vendor ships a fix — averaged 11 per month in that period, up from 8 per month in 2025.
•
August alone saw 22 — nearly double the monthly average.
•
In plain terms = attackers are finding and weaponising flaws at a pace that is visibly outrunning defenders.
03

How fast is an AI-driven attack in practice?

•
In August 2026, cybersecurity firm GreyNoise recorded an attack on a PaperCut system: the attacker escalated from initial breach to domain-admin access in minutes.
•
This means → AI-powered attack chains have drastically compressed the window defenders have to respond during lateral movement — hopping from one machine across a network.
•
Cloudflare also reported PHP exploitation attempts against its users in October. GTIG noted the two PHP flaws involved were discovered via large-language-model analysis; technical details were public before the patch dropped, so researchers and attackers started the race at the same time.
04

Is Google the only one sounding the alarm?

•
Anthropic's September 2026 report flagged the same shift: AI-assisted attacks have evolved into a multi-agent operational model that can automate most steps from discovery to data exfiltration.
•
In plain terms = operations that once required a coordinated team of human hackers can now be handled by a cluster of AI programs.
•
This reflects an industry-wide consensus forming — the threat is no longer one vendor's assessment.
05

What does this mean for enterprises?

•
The window between a vulnerability going public and being exploited is shrinking fast — automated attack pipelines make "patch not yet applied, attack already arrived" the new normal.
•
Whether an organisation can deploy patches faster than attackers can automate exploitation is becoming the defining metric of cyber-defence resilience.
•
This means → the priority for security spending needs to shift from "can we defend?" to "can we outpace the machines?"

市场有风险,内容仅供研究参考,不构成投资建议。

相关标的

4
  • CrowdStrikeCRWD.US-1.3%

    AI-automated attacks compress the patch window, forcing enterprises to accelerate procurement of automated endpoint detection and response solutions. CrowdStrike is the market leader in this segment.

  • Palo Alto NetworksPANW.US-1.8%

    AI-driven attack chains accelerate lateral movement, boosting demand for network perimeter and zero-trust defenses — a direct positive for leading firewall and SASE vendors.

  • Cybersecurity ETFBUG.ETF-1.8%

    AI-automated vulnerability exploitation raises overall cybersecurity spending expectations. This ETF covers the full cybersecurity spectrum with relatively good liquidity.

  • CloudflareNET.US-1.7%

    The report directly mentions Cloudflare users being targeted by PHP vulnerability exploits, validating the demand scenario for its WAF and application security products.