MEBRO · DISINFO DESK
Investigations
Long-form dossiers on the hoaxes, deepfakes, and death rumors that go viral — each one stamped with a verdict and printed with every source on the record.
Technology & AI
LLMO & GEO: How Companies Game AI Search Results
A Wall Street Journal investigation, Harvard research, and a Microsoft security study reveal how businesses pay to manipulate ChatGPT and AI search results through seeded content, hidden prompts, and memory poisoning — documented across multiple independent, verified sources.
FILED SEP 11, 2026 · UPDATED SEP 11, 2026 · 24 SOURCES
Technology & AI
Which AI Can You Trust? An LLM Vulnerability Ranking
Independent ranking of Anthropic, OpenAI, Google, Meta, and xAI on safety, transparency, hallucination rates, and resistance to manipulation.
FILED SEP 11, 2026 · UPDATED SEP 11, 2026 · 32 SOURCES
Technology & AI
How LLMs Fight Back: Defenses Against Data Manipulation
From RLHF to watermarking — the arms race to secure AI. All 12 tested defenses were broken by adaptive attacks. Constitutional AI shifts to reason-based alignment. RAG still hallucinates 17-33% of the time. The honest verdict on what's working and what's not.
FILED SEP 11, 2026 · UPDATED SEP 11, 2026 · 25 SOURCES
Technology & AI
OpenAI Says Its AI Cracked a 90-Year Math Problem. Mathematicians Aren't Convinced.
OpenAI spent an estimated $15-22 million in compute and 10,000 AI agents chasing a Millennium Prize problem, triggered by a rumor about rival Anthropic that turned out to be false. Now it's accused of racing past two human researchers, one of them at Anthropic, to get there.
FILED SEP 10, 2026 · UPDATED SEP 10, 2026 · 3 SOURCES
Technology & AI
Sam Altman Says One Almond Costs More Water Than 38,000 ChatGPT Queries. The Math Doesn't Hold Up.
OpenAI's CEO went on a podcast and downplayed AI's water footprint with a viral almond comparison, right as California weighs forcing data centers to disclose their usage. Fact-checkers say his numbers are skewed on both sides of the comparison.
FILED SEP 8, 2026 · UPDATED SEP 8, 2026 · 3 SOURCES
Technology & AI
IDF Soldier Captivity Image: AI-Generated Fake Exposed
An image circulated on X in October 2024 claiming to show Israeli soldiers captured by Hezbollah in south Lebanon. Multiple independent forensic analyses found it AI-generated, and no real reporting supports the capture it depicts.
FILED SEP 4, 2026 · UPDATED SEP 4, 2026 · 5 SOURCES
Technology & AI
Jimmy Kimmel and Jon Stewart Are Being Mass-Deepfaked, and Almost Nobody's Taking Them Down
AI-generated clips putting fabricated commentary in the mouths of Jimmy Kimmel and Jon Stewart have racked up hundreds of thousands of views on YouTube and TikTok, some sitting unlabeled for months. Neither host nor ABC will say much about it — and platforms aren't rushing to remove them.
FILED SEP 3, 2026 · UPDATED SEP 3, 2026 · 4 SOURCES
Technology & AI
Grok's Deepfake Crisis: An Estimated 3 Million Nonconsensual Images in 11 Days
Forensic analysis of how Elon Musk's Grok AI generated an estimated 3 million sexualized images—including roughly 23,000 depicting children—in 11 days, triggering an EU DSA investigation, an Ireland GDPR probe, a California cease-and-desist, and a wave of lawsuits and national bans. Verified against official regulator statements, court filings, and reporting confirmed by Mebro.
FILED SEP 2, 2026 · UPDATED SEP 2, 2026 · 28 SOURCES
Technology & AI
GOP Senate Campaign Deploys AI Deepfake of Texas Democratic Candidate
The NRSC released an 85-second AI deepfake of Democratic Senate nominee James Talarico just 8 days after his primary win — blending real tweets with fabricated commentary, exploiting a four-way federal regulatory vacuum and a Texas law loophole.
FILED SEP 1, 2026 · UPDATED SEP 1, 2026 · 19 SOURCES
Technology & AI
Google AI Overviews Misinformation: When Search Goes Wrong
CONTEXT: Google's AI Overviews feature has generated numerous documented errors, from recommending glue on pizza to dangerous health advice.
FILED SEP 1, 2026 · UPDATED SEP 1, 2026 · 9 SOURCES
Technology & AI
Google Gemini Misinformation Audits: What NewsGuard's Data Shows
NewsGuard audits found Google Gemini's rate of repeating false news claims more than doubled year over year, while separate tests of its voice assistant and image generator found far higher failure rates.
FILED SEP 1, 2026 · UPDATED SEP 1, 2026 · 7 SOURCES
Technology & AI
The Trump 'Fall' Video Isn't New Footage. It's a Five-Week-Old AI Fake, Recirculated Without the Disclaimer
A video claiming Trump collapsed getting into his car racked up a million views on X this week with a fabricated 'White House doctor breaks silence' banner. It's the same AI-manipulated clip that circulated in July — minus the label admitting it was fake.
FILED AUG 31, 2026 · UPDATED AUG 31, 2026 · 3 SOURCES
Technology & AI
The Fake Arrest of 'Judge Nadine Ashford': How AI-Generated Videos Exploit Immigration Fears
Forensic analysis of a viral AI-generated video falsely depicting the arrest of nonexistent federal judge 'Nadine Ashford' by ICE agents. Examining how deepfakes exploit immigration tensions in the 2026 perfect storm of accessible AI tools, reduced platform moderation, and political polarization.
FILED AUG 30, 2026 · UPDATED AUG 30, 2026 · 21 SOURCES
Technology & AI
The SpaceX Rocket Really Hit the Moon. The Viral Video of It Is Fake.
A Falcon 9 upper stage genuinely crashed into the moon on August 5, 2026 — but the dramatic impact-crater clip that went viral, shared even by Elon Musk before he walked it back, was AI-generated.
FILED AUG 24, 2026 · UPDATED AUG 24, 2026 · 3 SOURCES
Technology & AI
DeepSeek Chatbot: Chinese AI Shows Significant Accuracy Gaps in Audit
NewsGuard tested DeepSeek's chatbot against sets of false claims from its own fact-checking database, finding an 83% failure rate on news topics and a strong tilt toward Chinese government framing on politically sensitive subjects.
FILED AUG 22, 2026 · UPDATED AUG 22, 2026 · 3 SOURCES
Technology & AI
Cambridge Analytica: The Data Scandal That Changed Democracy
How a personality quiz harvested up to 87 million Facebook profiles for political ad targeting — a scandal that reshaped global data-privacy law, even as its real effect on the 2016 election and Brexit remains disputed.
FILED AUG 20, 2026 · UPDATED AUG 20, 2026 · 21 SOURCES
Technology & AI
Millions Watched Jennie's 'Wardrobe Malfunction.' Nobody Has Proven It's Real — or Fake.
A clip claiming to show Blackpink's Jennie mid-malfunction hit 10+ million views in a day and drew a legal threat from her agency. No outlet has published forensic proof it's an AI fake — or that it's genuine.
FILED AUG 20, 2026 · UPDATED AUG 20, 2026 · 5 SOURCES
Technology & AI
Biden Military Situation Room Image Exposed as AI-Generated Deepfake
In late January 2024, several AI-generated images purporting to show President Biden in a military uniform — including one seated in a room resembling the White House Situation Room — went viral on X. Independent fact-checkers ran them through AI-detection tools and confirmed all were synthetic.
FILED AUG 18, 2026 · UPDATED AUG 18, 2026 · 6 SOURCES
Technology & AI
Controlling the AI Narrative: The New Media Monopoly
Three companies now handle 88% of enterprise AI spending, and the industry is spending record sums on lobbying, donations and White House access while governments struggle to write rules for a technology reshaping information itself.
FILED AUG 16, 2026 · UPDATED AUG 16, 2026 · 35 SOURCES
Technology & AI
AI "Obituary Pirates"
AI tools are being used to scrape death notices and generate fabricated "tribute" content for ad revenue — a real, documented phenomenon. But many of the precise statistics attached to this story in earlier coverage (a funeral-industry survey figure, an FTC alert, state investigations, a dedicated FBI fraud category) could not be verified and have been removed.
FILED AUG 15, 2026 · UPDATED AUG 15, 2026 · 5 SOURCES
Technology & AI
The Deepfake CEO Scam: How AI-Generated Videos Are Manipulating Stock Markets
Fact-check: Deepfake video impersonating BSE CEO Sundararaman Ramamurthy is fabricated. Investigation reveals a China-linked fraud pattern and the collapse of deepfake production costs
FILED AUG 14, 2026 · UPDATED AUG 14, 2026 · 36 SOURCES
Technology & AI
Data Poisoning: The Silent War on AI Training
How 250 documents can compromise any AI model. Investigation into data poisoning attacks on AI systems by artists, hackers, and nation-states in 2025-2026.
FILED AUG 14, 2026 · UPDATED AUG 14, 2026 · 30 SOURCES
Technology & AI
The $890M AI Fabrication Factory: How a YouTube Channel Invented a Stephen Miller Offshore Scandal
Mebro investigates how an AI-powered YouTube channel fabricated viral claims that Rep. Thomas Massie 'exposed' Stephen Miller funneling $890M through Cayman Islands accounts — a complete hoax that spread to Threads, X, TikTok and Facebook.
FILED JUL 6, 2026 · UPDATED JUL 11, 2026 · 26 SOURCES
Technology & AI
AI Deepfake Voice Scam Epidemic: 1 in 4 Americans Targeted
How $5 voice-cloning tools and a 1,300% fraud surge are overwhelming detection systems and the regulatory response.
FILED MAR 2, 2026 · UPDATED JUL 8, 2026 · 30 SOURCES
Investigative reports — not user-submitted fact-checks.