Edition 7 · August 7, 2026
Google pulled its new Earth AI image tool within 24 hours after researchers used it to generate photorealistic fake satellite images of nuclear facilities and war zones, and its own SynthID watermark and Gemini failed to flag them as fake. That’s the week in one story: the fakes are getting good enough to beat the detectors built to catch them, aimed now at the satellite imagery, elections, and disaster reporting people treat as ground truth. Around it: Russia ramping up disinformation before German and French elections, a social-media campaign that pushed tens of thousands of migrants toward Ceuta, and Meta reportedly earning $7 billion a year from scam ads.
Deepfakes & Detection
- Google withdrew its Earth AI image generator within a day (Tech Times) after researchers made fake satellite imagery of nuclear sites that its own SynthID watermark and Gemini couldn’t reliably identify; the company is rewriting its policies (SCMP).
- A UK investigation found 276,149 sexual deepfakes on a single site, 96% depicting women (New Atlas), spurring new detection tools.
- AI deepfakes are already in this US election cycle (Fast Company): a Michigan Senate video casting a Democrat as Odysseus, and AI attack ads in the Massachusetts governor’s race.
- After Japan’s deadly Kumamoto earthquake, conspiracy posts claimed it was “man-made” (The Independent), amplified with AI images across languages.
Election Interference
- Security sources say Russia is stepping up disinformation before German state elections (The Star), targeting every major party except the far-right AfD with fabricated claims about local politicians.
- France’s SGDSN warned that ex-PM Gabriel Attal was targeted by a Russia-linked operation (RFI) ahead of the 2027 vote, following similar alerts about other candidates.
- Russia’s embassy in Switzerland dismissed local “interference” reporting (TASS) ahead of a neutrality referendum, the standard denial worth logging alongside the allegations.
The Ceuta Disinformation Surge
- Morocco confirmed an online campaign across TikTok, WhatsApp, Facebook, and Instagram (Jerusalem Post) mobilized people to cross into Spain’s Ceuta enclave; 50,000+ crossed and 48,000+ returned within 48 hours.
- The posts falsely claimed Spain had relaxed migration rules (Euronews), mixing old videos and AI images; Spain’s Supreme Court then okayed returning migrants (Al Jazeera), a contested call on non-refoulement.
AI as Attack Surface
- A new prompt-injection technique spreads through “Ask AI” buttons (The Hacker News) on commercial sites, silently altering an assistant’s memory with no malware needed.
- OpenAI disrupted a Cambodia-based scam network (The Hacker News) that used ChatGPT to run investment, romance, and law-enforcement-impersonation schemes.
- Oxford’s Internet Institute launched InfoOps Bench, a live AI-safety benchmark (Tech Times) scoring whether models resist generating disinformation.
Platform Accountability
- A 2025 report says Meta earned about $7 billion a year from scam ads (NDTV), only banning advertisers when 95%+ certain of fraud.
- The EU AI Act’s deepfake-labeling and chatbot-disclosure rules took effect August 2 (Analytics Insight), with fines for non-compliance.
- Ad giant Publicis bought data broker LiveRamp for $2.2B (Breitbart), a deal critics tie to online ad-blacklisting debates.
India’s Deepfake Cases
- India is a hotspot this week, though partly because it produces so much English-language coverage. The Bombay High Court ordered Meta, X, and Google to remove deepfakes of minister Nitin Gadkari (Economic Times), an early liability precedent.
- Police filed cases over deepfakes of PM Modi and minister Piyush Goyal (Asianet), and are probing 180 accounts and a “foreign bot network” (News18) behind protest content. The foreign-attribution is a government claim, not yet independently verified.
Crime & Cyber
- Eleven countries jointly warned that North Korean IT workers are infiltrating Western firms (Bluewin) under false identities to steal data and funnel wages to the regime.
- Researchers tied a wave of hedge-fund cyberattacks to the UNC6671/BlackFile extortion group (BleepingComputer).
- Malaysia’s Star Media flagged a fake article impersonating its brand (The Star) to spread false claims about an anti-corruption official.
Conspiracy Watch
- Monitors flagged a surge of antisemitic conspiracy posts blaming Jews for European wildfires (Jerusalem Post), with near-identical narratives across languages and 2M+ views on one thread.
- A five-year study found persistent media skepticism shapes common vaccine-conspiracy beliefs (News-Medical), separating institutional distrust from belief in implausible claims.
Media & Literacy
- Kenya’s fact-checking network hit a “timing trap” (The Conversation): misinformation spread faster than checkers could verify it.
- Academics warn the Philippines’ anti-fake-news bill could harm civil liberties (Inquirer) more than it curbs disinformation.
- Editorial: rising news avoidance is leaving societies less informed (Ara), feeding authoritarian populism where journalism declines.
Try This — Bad Aura
- A recruiter reading your feed isn’t looking for your best posts. They’re looking for reasons to pass. A scammer isn’t either: they’re looking for an opening.
- Bad Aura, from Aunoo, connects to your account and reads your public posts back to you through their eyes: what a recruiter flags for (un)employability, and what a scammer sees to work with.
- Same posts, two verdicts you never get to see. Now you can.
- See how they see you →
About Bad Information
A weekly read on disinformation, coordinated influence operations, and media literacy. Researched and fact-checked with Aunoo.
Powered by Aunoo — news intelligence with trust built in. Aunoo scores every article on five trust signals before you even click: source bias, factual-reporting rating, AI-content detection, disinformation-network flags, and a quality score. You see how far to trust a story, not just what it claims.
Try it, it’s free → saas.aunoo.ai
AI Transparency
This newsletter is produced with AI. Aunoo uses automated systems to gather, score, and draft the content, with human review before publication. In line with the EU AI Act’s transparency obligations (Article 50), which took effect August 2, 2026, we disclose that parts of this edition were generated or assisted by artificial intelligence. Every claim links to its original source — please verify against those before reuse.
