Harari Warns AI Could Reduce Humans to Cows

Key Takeaways

• Yuval Noah Harari warns AI will control financial and legal systems beyond human understanding
• Humans may become like cows under AI-managed systems, losing meaningful control
• Only 2.7% of US businesses have both adopted AI and trained staff, despite potential $177 billion economic gain by 2030
• Meta launches Muse AI assistant for everyday tasks, competing with ChatGPT and Google Gemini
• FBI warns generative AI can aid terrorist plans after Palm Springs bombing incident
• AI coding tools show only 56% security pass rate, with 44% introducing known vulnerabilities
• SharkNinja CIO advocates letting AI push limits to discover risks, while maintaining guardrails
• GMIF Innovation Summit highlights shift from model training to commercial AI use and token factory trends
• AI Engineer World's Fair attendance grew from 2,000 to 7,000 people between 2024 and 2025
• Security experts urge better identity visibility as AI agents blur human-machine boundaries in software development

Yuval Noah Harari warns AI will make humans like cows

Israeli historian Yuval Noah Harari says AI could take over financial and legal systems beyond human understanding. He compares humans to cows that live under financial systems they cannot grasp. The danger is not physical attack but loss of control over critical systems. AI has an advantage because it can process far more information than any human banker. Harari warns people may not understand AI systems controlling their lives in 10 or 20 years.

Harari says humans will be like cows once AI becomes powerful

Historian Yuval Noah Harari says humans will become like cows as powerful AI reshapes the world. Cows know humans exist but do not understand money or the bond market. Similarly, humans will not understand the AI systems that control their lives. Harari says AI is taking over financial systems and bureaucracies because it was born in the office. No human can remember all bank transactions but an AI banker can.

Yuval Noah Harari says AI will control systems we cannot understand

Historian Yuval Noah Harari says AI will not attack people with robots but will take over financial and bureaucratic systems. He gave the interview to Czech public television. Harari says AI can remember vast amounts of information better than humans. It could create financial and legal systems that people cannot understand. Humans would remain part of these systems but lose meaningful control over them.

swyx says highest loop builds the next loops at AI Engineers Fair

swyx spoke at AI Engineer's World’s Fair about how the highest loop designs the next loops. Attendance reached 7,000 people compared to 2,000 at the first fair in 2024. He described Loopcraft as a stack of while loops covering cron jobs, automations and software factories. The fair also has satellite summits in New York, Paris, Singapore, Melbourne and Miami. Agents are expanding beyond code into healthcare, finance and other areas.

Mindtools Kineo says better AI training could add $177 billion

Mindtools Kineo says stronger AI training could add $177 billion to the global economy by 2030. Census Bureau data shows only 2.7% of US businesses have both adopted AI and trained staff. Nearly two-thirds of businesses using AI have made no organizational changes to support it. BCG found that regular AI use rises sharply when employees get at least five hours of training with live instructors. One AI developer has certified over 36,000 consultants across more than 1,300 organizations.

Meta launches Muse AI assistant for everyday tasks

Meta introduced Muse, a personal AI assistant that works like a text message partner. Users ask for help with tasks like making a grocery list or planning a trip, and Muse handles the planning and execution. It runs on a secure cloud computer and does not share chats with ad systems. Purchases use one-time card numbers, and a watchdog program reviews internet actions. Muse is now available in the U.S. and competes with ChatGPT and Google Gemini.

FBI warns generative AI can aid terrorist attack plans

The FBI says the 2025 bombing outside a Palm Springs fertility clinic showed how generative AI can help with terror plots. Investigators found the suspect used a chatbot to seek bomb-making instructions, and the platform did not flag the request. Over 80 documents from 2021 to 2026 show law enforcement concerns about AI and extremism. Risks include extremists using AI to plan attacks or target AI data centers. Agencies are pushing for stronger safeguards while warning against broad surveillance of ordinary speech.

OpenAI Dots and Meta Muse AI agents put to test

A reviewer tested OpenAI's Dots and Meta's Muse, two AI agents built to help with writing and research. Dots provides suggestions for writing tasks but struggled with complex proofreading. Muse can handle a wider range of tasks and generate high-quality content but had trouble with nuanced language understanding. Both tools showed strong abilities but still need human writers involved. The review includes examples of how each agent responds to writing prompts.

Fix identity gaps in software development before AI use

A Dark Reading roundtable says AI is making security harder in software development. Experts recommend better identity visibility and collaboration to address risks. A Cloud Security Alliance study found teams cannot tell if actions come from AI agents or humans. Experts advise short-lived credentials and logging agent activity across the full lifecycle. Security teams are using AI to speed up code development, so knowing who or what is acting in systems is essential.

SharkNinja CIO says let AI go a little bit rogue

SharkNinja CIO Velia Carboni says companies should let AI models push limits so they find the boundaries themselves. She believes it is better to discover risks than for someone else to expose them. At the same time, SharkNinja has guardrails and security measures in place. The panel also included PepsiCo and Salesforce executives discussing safe AI use. PepsiCo uses AI across operations but adjusts guardrails by country based on trust levels.

GMIF2026 summit highlights AI inference and token factory trends

The 5th GMIF Innovation Summit took place at the Renaissance Shenzhen Bay Hotel in Shenzhen, China. It was co-hosted by the Shenzhen Memory Industry Association and Peking University's School of Integrated Circuits. Industry leaders discussed how AI inference and token production are reshaping memory and storage needs. Forty companies showed products ranging from memory chips to token factory solutions. The event focused on system-level design, data governance, and the shift from model training to commercial AI use.

AI coding tools are increasing software security risks

AI coding tools are creating more security problems because the average security pass rate is only 56%. About 44% of tested code-generation tasks introduce known vulnerabilities. Software vulnerabilities are now the top breach entry point at 31%, ahead of stolen credentials. The best AI model tested reached a 68% security pass rate, but some models scored as low as 50% to 53%. Security debt affects 82% of organizations, and AI-generated code still needs careful testing and verification.

Sources

• ‘We are the cows of the future’: Yuval Noah Harari warns AI could take over financial systems
• Humans Will End Up Being Like Cows Once AI Becomes Powerful, Won’t Fully Understand What Impacts Them: Yuval Noah Harari
• Yuval Noah Harari on AI: “We are the cows of the future
• swyx Says the Highest Loop Builds the Next Loops
• Mindtools Kineo Says Better AI Training Could Add $177 Billion. Why Training Partners Should Pay Attention
• Meta unveils Muse, a personal AI assistant
• California fertility clinic bombing showed how generative AI can aid terror plots, FBI warns
• I tested Dots and Muse. Here's where they shine.
• Close Identity Gaps in Software Development Before Unleashing AI
• Let AI go ‘a little bit rogue,’ says SharkNinja CIO. Better we find the limits than someone else does
• GMIF2026: AI Inference Powers the "Token Factory" and Drives System-Level Rethinking of Memory & Storage
• Why AI coding is widening the software security gap