Iranian operatives deployed seven artificial intelligence-generated journalist personas, including a fictitious freelancer named Ervin B. Hoskins, to plant nearly 100 anti-American articles across roughly a dozen international online publications. OpenAI reported that the covert influence campaign utilized ChatGPT to generate story pitches, draft provocative op-eds, and amplify content across social media platforms to shape foreign public opinion.
The operation specifically targeted small to medium-sized US media outlets, framing Washington as the aggressor in an ongoing seven-month conflict while portraying Israel unfavorably. This disclosure follows similar findings by Anthropic regarding state-backed exploitation of LLMs for cognitive warfare and dissident surveillance.
The integration of large language models into covert influence campaigns scales the generation of synthetic authority by automating the creation of verifiable digital footprints. By deploying multi-persona personas equipped with AI-generated imagery and interconnected social networks, state actors bypass traditional editorial gatekeeping to infiltrate niche ideological publications. This operational method transforms resource-heavy disinformation into an automated, highly adaptable asset that exploits the vulnerabilities of decentralized media distribution systems.
No comments:
Post a Comment