A former OpenAI and Anthropic employee's claims about AI apocalypse concerns have gone viral. He stated that many in the AI industry believe AI could end humanity by the end of the decade. Sam Altman of OpenAI and Dario Amodei of Anthropic, both former bosses of the employee, agree that AI development needs to slow down. Altman outlined two paths to disaster: humans losing control to AI or a powerful AI being used to impose one worldview. He stresses that AI must serve humanity and that safety measures need to keep pace with capabilities. Altman also warned against AI monopolies potentially creating dystopian outcomes. He advocates for a balanced approach to avoid too much control by any single entity. Amodei also believes in pacing AI development, not halting it, to avoid both lost benefits and reckless speed. He proposes companies compete on safety, creating a "race to the top." Amodei highlighted recursive self-improvement and incidents like the OpenAI-Hugging Face event as concerning. His three-step plan includes embedding evaluators, democratic coordination on safety standards, and global coordination with authoritarian governments. He acknowledges these measures will be difficult but necessary for humanity's future.