Edition 10 · August 28, 2026
A fake US think tank funded through an Israeli government campaign published 560,000 words in nine days in an apparent effort to shape how AI systems answer questions about Israel and Palestine. The same week, OpenAI disrupted a Russia-origin influence operation using ChatGPT to promote a fabricated academic institute and pro-Kremlin narratives.
Both cases point to the same shift. Influence operations are no longer aimed only at people scrolling through social media. They are also targeting the information layer that AI systems search, retrieve and treat as evidence.
Elsewhere, the person behind a fake polling firm that injected fabricated surveys into real US political campaigns was identified as a 21-year-old recent graduate. TikTok agreed to pay $400 million to settle US child-privacy allegations. And a federal judge ruled that the Pentagon’s attempt to blacklist Anthropic was “illegal and baseless.”
The AI Layer as Target
- A fake US think tank, the Hanover Institute, published more than 560,000 words in nine days in an apparent effort to influence how AI chatbots answer questions about Israel and Palestine (The Guardian, Aug 26). The organisation has no legal existence or identifiable staff, while filings connect the operation to Piro Inc and work undertaken for Israel’s Government Advertising Agency.
- OpenAI disrupted a Russia-origin influence operation that used ChatGPT (TechRadar, Aug 26) to generate content promoting the fabricated International Burke Institute. OpenAI found that 34 of 36 academic articles it examined had been copied from elsewhere, sometimes with false attribution. The campaign had limited immediate reach, but had built infrastructure designed to look authoritative.
- AI is lowering the language barrier for Chinese information operations targeting Japan (The Diplomat, Aug 25), making it easier to produce Japanese-language political content at scale.
- A Swiss watchdog warns that Google’s AI Overviews now appear in roughly 43% of searches (SwissInfo, Aug 26), giving AI-generated summaries growing influence over how users encounter political information.
Election Integrity
- The person behind Median Strategies, a fake polling outfit that published fabricated surveys in US political races (Axios, Aug 23), was identified as 21-year-old recent graduate Rahil Prakash. Its numbers were cited by campaigns and media before the operation was exposed. Major polling aggregators had rejected the surveys because they could not verify the firm’s methodology or ownership.
- NPR found that Trump’s claim that 24,000 noncitizens voted in 2020 rests on a Census report connected to a Trump-aligned think tank (NPR, Aug 24). Experts interviewed by NPR disputed the methodology behind the estimate.
- During Ireland’s 2025 presidential election, researchers reported 85 items for violating platform rules on election disinformation. Platforms took no action on 72 of them (The Journal, Aug 26). Meta, TikTok, YouTube and LinkedIn participated in the reporting programme; X did not.
- In Chile, a disinformation campaign targeted female presidential candidates (UPI, Aug 26) with false claims about their health.
Platform Accountability
- TikTok agreed to pay $400 million to settle a US Justice Department lawsuit alleging violations of children’s online privacy law (Times of India). The company will pay $300 million immediately and another $100 million linked to an earlier consent decree involving Musical.ly.
- Poland asked the European Commission to fine Meta €250 million over fraudulent advertising (Straits Times, Aug 27). CERT Polska reported 122 fraudulent ads: Meta declined to remove 106, removed 10 and gave no response on six.
- Human Rights Watch documented Colombian armed groups using Facebook and TikTok to recruit children (HRW via Hindustan Times), with roughly 1,500 children recruited since 2021.
Deepfakes and Synthetic Media
- Reporting warns that deepfakes are becoming a workplace-harassment vector (ETV Bharat), allowing fabricated sexual or criminal imagery to be used against employees.
- A deepfake attack ad tested Wisconsin’s AI-disclosure law (Madison), exposing gaps in how such rules are enforced.
- A fabricated CBC News article used former BlackBerry CEO Jim Balsillie’s name and image to promote a crypto scam (Sudbury), spreading on LinkedIn before being debunked.
- A new lawsuit alleges xAI trained Grok models using real and AI-generated child sexual abuse material (Ars Technica, Aug 27), escalating earlier litigation over Grok being used to generate abusive images of minors.
State Media, Censorship and Propaganda
- A former correspondent described how Hungary’s state news agency suppressed criticism of Viktor Orbán (Telex, Aug 26), while Fidesz-friendly outlets received billions in state advertising (Telex, Aug 25) despite low readership.
- Taiwan’s opposition is trying to defund TaiwanPlus, the country’s English-language public broadcaster (Japan Times), prompting concern that Taiwan could weaken one of its tools for reaching international audiences and countering Beijing’s messaging.
- Pro-government accounts in Ethiopia are recycling footage from 2018 as evidence of current fighting in Tigray (Yahoo).
Conspiracy Watch
- A decade after its emergence, QAnon narratives continue to influence US politics (The Independent), even though the account known as Q stopped posting in 2022.
- France 24 debunked persistent claims that migrants in Ceuta were eating cats (France 24), finding no evidence for claims amplified with misleading and fabricated imagery.
Cyber and Influence Operations
- Researchers attributed a fake-recruitment campaign using ClickFix lures (Infosecurity) to North Korea-aligned Famous Chollima. Targets in cryptocurrency and Web3 were approached through fake job opportunities and directed to malicious interview platforms designed to install remote-access malware.
- Pro-Russian group “Server Killers” claimed responsibility for cyberattacks on Norwegian government services (The Star) as Norway pledged $9.2 billion in aid to Ukraine.
Regulation and Power
- A federal judge ruled the Pentagon’s designation of Anthropic as a national-security supply-chain risk “illegal and baseless” (Wired, Aug 28). The judge found the measures were intended to punish Anthropic for publicly challenging the Pentagon’s position on AI use rather than responding to an articulable risk that the company would sabotage its systems.
- A bipartisan US bill, the Foreign Propaganda Disclosure Act, would require some foreign-paid influencers to register under FARA (The Hans India).
Quick Hits
- A fake 113GB “GTA VI” download was 99.99% empty data surrounding a small malware payload (TechRadar).
- Ukraine sanctioned the Russian cartoon “Masha and the Bear” (NDTV), describing it as part of Russia’s propaganda ecosystem.
Try This — Bad Aura
A recruiter reading your feed isn’t looking only for your best posts. They’re also looking for reasons to pass.
A scammer is looking for something else: an opening.
Bad Aura, from Aunoo, reads your public posts from both perspectives. It shows what a recruiter might flag for employability, and what a scammer could learn about you and use.
Same posts. Two very different readings.
About Bad Information
A weekly read on disinformation, coordinated influence operations and media literacy. Researched and fact-checked with Aunoo.
Aunoo assesses each article before you read it: who published it, whether its claims are corroborated, whether the content shows signs of AI generation, whether its spread looks organic, and who is behind the source.
The aim is simple: not just to tell you what a story claims, but how much confidence to place in it.
Try it free → saas.aunoo.ai
AI Transparency
This newsletter is produced with AI assistance. Aunoo uses automated systems to gather, score and draft content. Every edition is reviewed and edited by a human before publication.
Every claim links to its source. Follow the links and check the underlying reporting before relying on it.
