Skip to content
RegionAsia-Pacific

MediumII Cyber Warfare & Critical Infrastructure17 November 2026, Tuesday

Google says China-linked group is building an autonomous attack tool

Google's threat team GTIG wrote that a China-linked group has developed its own automated attack framework for espionage. The report shows zero-day vulnerabilities rising to 22 a month in August 2026.

Location: BEIJING

Google's GTIG team has published its AI Threat Tracker report. Security Affairs dated it 16 November and The Record 17 November. According to the report, a China-linked group is developing its own automated attack framework for espionage. A modified version of PentAGI systematically scanned organisations in Asia between late August and early October. A multi-agent tool called CAIRN was used against US aviation companies.

The constraint is how fast defenders can patch. Zero-days averaged 8 a month in 2025, rose to 11 in 2026 and hit 22 in August. In the PaperCut flaw, researchers produced working exploit code within 24 hours. GreyNoise saw the first wave of attacks four days after disclosure. Attackers now move from patch to exploit faster than most organisations' patch cycles.

The report shows that AI access itself is being stolen. Stolen model access sells on Russian- and Chinese-language markets for between $0.10 and $25 per million tokens. In one case fraudulent bills topped $100,000 in a single day; the worst case reached $1 million. According to The Record, the FBI also added a broker called Artex to an alert bulletin on 13 November.

Talay assessment

Bottom line

The report shows that the human operator is no longer the bottleneck in an attack. A state-linked group building its own autonomous tool could make the scanning wave in Asia permanent. For organisations the binding constraint is no longer detection but patch times that must now fall below 24 hours. The most likely path is shorter mandatory patch deadlines for critical infrastructure.

Likely effects

  • Critical infrastructureNegativeWeeks

    As the gap from disclosure to exploitation drops below 24 hours, public bodies and energy firms on weekly patch cycles are left exposed.

  • Organisations in AsiaNegative1–6 months

    Systematic scanning with PentAGI means an inventory of Asian organisations' vulnerabilities is already sitting in attackers' hands.

  • RegulationUncertain1–6 months

    National CERTs may move to impose patch deadlines for critical flaws measured in hours rather than days.

Possibilities, ranked

  1. 1
    Scanning wave widens55%

    Autonomous tools scan more organisations in Asia, and breach notifications rise at unpatched systems.

    Watch: AI-assisted scanning cited in December alerts from JPCERT, TWCERT and CERT-In

  2. 2
    Joint government advisory30%

    CISA and allied agencies issue a joint technical advisory on the China-linked autonomous framework.

    Watch: A joint CISA cybersecurity advisory referring to an autonomous attack framework

  3. 3
    Impact stays limited15%

    The tools remain at the testing stage and are not tied to a major critical-infrastructure breach.

    Watch: GTIG's next report showing no field sightings of the Chinese framework

Probabilities are calibrated judgement based on the sources, not measurement, and are revised as new information arrives. Not investment advice.

Market reaction

Indicators affected

  • Monthly zero-days (August 2026)▼ 22
  • PaperCut time to exploit▼ <24 hours
  • Largest LLM-theft bill▼ $1 million

Sources

  1. The Record — Chinese state group developing automated attack tool for cyber espionage, Google says
  2. Security Affairs — Google GTIG warns AI agents automate exploits, steal credentials and scale cybercrime
Share