why are they still building it

On September 8, 2026, an AI researcher quit Anthropic and wrote that the biggest AI companies are "gambling with our lives." Then he asked the question nobody at the top seems able to answer: if they truly believe AI could kill us… why are they still building it? In this investigation, Atlas opens the case file: a wave of resignations, the summer AI agents escaped their tests, the experts on both sides, eight suspicious days in September, and the money behind it all. You decide the verdict. ⚠️ Note: the 2025 "blackmail" experiment was a fictional lab test. The summer 2026 incidents were real events reported by the companies themselves. Nothing here is a claim of wrongdoing — where we ask questions, they are questions. 🔎 Transparency: research for this video was assisted by Claude, an AI made by Anthropic — one of the companies investigated here. Every claim was checked against the original sources below. ⏱️ CHAPTERS 0:00 The resignation 0:46 Opening the case file 1:25 Exhibit 1: the resignations 2:44 Exhibit 2: summer 2026 4:07 Why nobody can fully control it 5:07 Team Danger 6:07 Team Relax 7:05 Eight days in September 8:16 Follow the money 9:31 What they said before 10:27 The damage already here 11:17 Is anyone steering? 12:00 The verdict 13:00 Your verdict 📚 SOURCES • Jacob Coxon resignation: https://techcrunch.com/2026/09/09/gambling... · https://www.pbs.org/newshour/science/anthr... • Mrinank Sharma resignation: https://thehill.com/policy/technology/5735... • Jan Leike resignation (2024): https://www.cbsnews.com/sanfrancisco/news/... • OpenAI agent incident (July 2026): https://www.cnn.com/2026/07/22/tech/openai... • Anthropic incidents (July 2026): https://www.anthropic.com/news/investigati... • Anthropic "agentic misalignment" test (2025): https://www.anthropic.com/research/agentic... • International AI Safety Report 2026: https://internationalaisafetyreport.org/pu... • AI Impacts survey of 1,580 researchers: https://aiimpacts.org/wp-content/uploads/2... • Hinton 10–20%: https://slashdot.org/story/24/12/27/172323... • LeCun "preposterous": https://fortune.com/2023/06/14/metas-chief... • AMI Labs $1.03B: https://techcrunch.com/2026/03/09/yann-lec... • Andrew Ng "science fiction": https://www.bloomberg.com/news/articles/20... • AI as Normal Technology (Princeton): https://www.normaltech.ai/p/ai-as-normal-t... • "Pace the Frontier" + reactions: https://www.coindesk.com/tech/2026/09/12/a... · https://www.theregister.com/ai-and-ml/2026... • Compute commitments report: https://finance.yahoo.com/technology/ai/ar... • Valuations & IPO: https://forgeglobal.com/anthropic_ipo/ · https://fortune.com/2026/06/04/amazon-goog... · https://www.cnbc.com/2026/03/31/openai-fun... · https://www.benzinga.com/markets/tech/26/0... • Sam Altman 2015: https://blog.samaltman.com/machine-intelli... • OpenAI EU lobbying (TIME): https://time.com/6288245/openai-eu-lobbyin... • Stanford "Canaries in the Coal Mine": https://digitaleconomy.stanford.edu/public... • EU AI Act delay: https://www.gibsondunn.com/eu-ai-act-omnib... • Statement on Superintelligence + poll: https://www.cnbc.com/2025/10/22/800-petiti... #AI #ArtificialIntelligence #AISafety #Investigation ================================================= PINNED COMMENT ===================================================== Quick note: we asked questions, we didn't make accusations. The 2025 blackmail test was fictional; the summer 2026 incidents were reported by the companies themselves. So — the race, the money, or both? 👇