AI pushed Awami League, smeared rivals, echoed India lines in Bangladesh

DHAKA (MNTV) — An operator in Bangladesh used the Claude artificial-intelligence chatbot to run an automated network churning out fabricated Bengali-language “news” that promoted the Awami League party and smeared its rivals, according to a report from the AI’s US maker, Anthropic.

The case is one of nine influence operations detailed in Anthropic’s threat-intelligence report, published Sept. 10, examining misuse of its models between December 2025 and August 2026.

As with the report’s other cases, the findings reflect Anthropic’s own investigation; the company said it removed the network and could not confirm it reached a wide audience. The claims could not be independently verified.

According to the report, an operator based in Gaibandha District, in northern Bangladesh, rotated through 29 Claude accounts to evade detection and used a custom program — named “fake_news_3.py” — to generate content in fixed batches: at least 1,500 fabricated headlines, 300 false narratives, and 1,500 image prompts.

The material was designed to feed a continuous cycle of Facebook, YouTube, and TikTok livestreams aimed at rural Awami League supporters with limited literacy.

In internal notes, the operator described the output as “fake news” calibrated to be “hot and aggressive” and simple enough “so even village people understand,” and wrote that “no one knows the news is fake.”

Because Claude cannot generate images, Anthropic said, the prompts were probably exported to other AI systems to produce the accompanying visuals.

The content was uniformly favorable to the Awami League — which has been out of power since the July 2024 uprising that ousted former Prime Minister Sheikh Hasina, meaning the network worked for an opposition party rather than the government — and attacked its opponents, including the Bangladesh Nationalist Party (BNP), the Jamaat-e-Islami party, the National Citizens Committee, the interim government, and student protest leaders.

The network ran fabricated smear campaigns accusing those opponents of being “foreign agents” and of advancing “Taliban-style governance,” the report said.

Anthropic stressed that despite the content’s alignment with Awami League interests, it “found no proof that the party itself was involved in directing or funding” the operation.

Pro-India propaganda

The report also noted that some of the network’s narratives “aligned with pro-Indian geopolitical interests” — a notable detail given India’s close ties to the Hasina government and Bangladeshi sensitivities over Indian influence — but added that it “found no evidence of any state direction or funding behind this activity,” leaving the alignment unexplained.

Anthropic assessed the operation’s reach at a mid-level “Category Three,” saying videos matching its output appeared on multiple Bangladesh-focused social-media accounts, though it could not confirm they spread more widely.

The case illustrates a pattern the report describes across several countries: AI letting an operator cheaply automate the work of a propaganda office, tailoring emotionally charged falsehoods to specific audiences.

For Bangladesh — where political tensions have run high since the 2024 uprising, with national elections anticipated — advocates warn such tools could deepen an already polarized information environment, particularly among rural voters least able to identify fabricated content.