OpenAI reveals a Chinese influence operation that leveraged ChatGPT as a diary and other tools to manage a large network of fake accounts, including impersonating officials and fabricating obituaries and forged documents to harass critics.

A planning session in mid-October 2025 explored a covert influence campaign targeting Japanese politician Sanae Takaichi, though the model refused to assist with that plan.

Another effort attempted to draft a strategy to undermine Japan’s prime minister by stoking online anger over US tariffs; ChatGPT refused, yet similar narratives later surfaced online.

Experts, led by Ben Nimmo, stress the need for vigilance, better detection, international cooperation, and robust AI governance to counter these evolving threats.

Environment-layer risks in AI resemble traditional IT threats but are amplified by scale and complexity of AI workloads, making attacks harder to detect.

Potential business impacts include higher security spending for AI vendors, stricter data governance in MLOps, and revised enterprise clauses on data residency, telemetry minimization, and bot mitigation.

A broad takeaway is that both state and non-state actors are increasingly using AI in influence operations, necessitating ongoing monitoring by platforms and researchers.

AI investment and competitive dynamics are shifting, with significant private-sector stakes amid concerns about IP theft and cyber threats in the sector.

Google is pursuing proactive defenses with better detection, asset takedowns, safety measures, and industry collaboration; tools like Big Sleep and CodeMender illustrate proactive vulnerability discovery and remediation.

Threat actors show growing interest in agentic AI concepts, attempting to prompt advanced models with specialized personas and automate vulnerability analysis, C2 development, and data exfiltration, though real-world autonomous agents remain limited.

The piece notes potential economic and social repercussions, including AI’s influence on democracy, calls for AI literacy, and shifts in geopolitical dynamics in the Asia-Pacific.

The OpenAI report arrives amid broader industry debates on safety versus innovation, highlighted by Anthropic’s relaxation of some guardrails under external pressure.