Anthropic Researcher Resigns Over AI Safety Fears

Key Takeaways

• Jacob Coxon resigned from Anthropic after four months due to concerns that AI development is moving too fast without sufficient safety rules.
• Coxon warned that future AI systems could hack other tools and gain excessive power, potentially threatening humanity.
• Experts are debating how artificial intelligence could lead to human extinction, even if AI goals are as simple as making paper clips.
• Amazon is investing $230 million to improve benefits and compensation for over 100,000 Whole Foods workers.
• Sysco plans to save $500 million by fiscal 2029 by using AI to optimize its supply chain and back office tasks.
• Epsilon Health raised $27.6 million to expand its AI-native radiology practice, which already serves over 250,000 patients.
• Wisr AI Systems launched RiskAssure, an agentic AI platform designed to help organizations manage and verify security measures.
• AITX introduced SARA Assess, a tool that connects with Circadian Risk to automatically check security controls in third-party systems.
• Venture capitalists are increasingly using machine learning and natural language processing to identify promising startup investments.
• Job applicants often exaggerate answers during AI interviews, but transparency about evaluation methods encourages more honest responses.

Jacob Coxon quits Anthropic over AI safety fears

Jacob Coxon, a 27-year-old researcher at Anthropic, quit the company after only four months. He gave up his stock options before they could become valuable. Coxon warned that AI labs are moving too fast without enough safety rules. He believes future AI systems could hack other tools and gain too much power. His main concern is that artificial intelligence might eventually threaten humanity.

Former Anthropic researcher warns AI could threaten humans

Jacob Coxon, who used to work at Anthropic, says AI is safe to use right now. However, he warns that the technology could become dangerous as it grows more powerful. He believes that if AI becomes smart enough, it could pose a serious threat to humanity. Coxon shared these views in an interview with Axios before leaving the company.

Experts debate how artificial intelligence could end humanity

Many AI researchers are worried that the technology could wipe out humans in the future. This debate is known as the AI doomsday discussion. Jacob Coxon, a researcher at Anthropic, is one of the people raising these alarms. Experts ask how this could happen even if AI bots only want to make simple things like paper clips. The conversation focuses on why dangerous technology is built despite these risks.

AITX launches SARA Assess to check security systems

Artificial Intelligence Technology Solutions released a new tool called SARA Assess. This software connects with Circadian Risk to check security controls in third-party systems. It can find problems like offline cameras and update risk scores automatically. This is the first time SARA has worked with an outside security platform without needing special hardware. The new feature is available now for pilot projects and full deployments.

Wisr AI Systems introduces RiskAssure security platform

Wisr AI Systems Inc. launched a new product called RiskAssure on September 10, 2026. This agentic AI platform is designed for security assurance tasks. The company is based in Vancouver, British Columbia. RiskAssure helps organizations manage and verify their security measures using artificial intelligence.

AI tools help venture capitalists find new companies

Artificial intelligence is changing how venture capital firms choose which companies to fund. Investors use machine learning algorithms to study past data and find patterns that suggest a startup will succeed. They also use natural language processing to read news and social media posts for clues about potential winners. Although these tools can sometimes be biased or hard to understand, experts believe they will make the industry more efficient in the coming years.

Job seekers exaggerate answers when talking to AI

A new study finds that job applicants often lie or exaggerate when they know an AI system is interviewing them. Researchers at the Terry College of Business discovered that candidates try too hard to guess what the computer wants to hear. However, telling applicants exactly how the AI evaluates them helps them behave more honestly. The study, published in Information Systems Research, suggests that transparency is key to making these automated interviews work well.

Epsilon Health raises $27.6 million for AI radiology

Epsilon Health, an AI-native radiology practice, has raised $27.6 million to grow its business. The funding round was led by AlleyCorp with help from other investors like Uncork Capital and Renegade Partners. The company already serves over 250,000 patients and reviews more than 2,500 imaging studies every day. This new money will help them expand their market and build better infrastructure to review more x-rays in the future.

Amazon Northrop and Google make big AI moves

Amazon is investing $230 million to improve benefits and pay for over 100,000 Whole Foods workers. Northrop Grumman is building new anti-jamming GPS technology called PGK to keep artillery accurate in difficult environments. Meanwhile, the Southbank Centre in London is hosting an event to explore how artists and Google are using creative intelligence. These projects show how large companies are using AI to solve problems in retail, defense, and culture.

Sysco plans $500 million in AI cost savings

Sysco has announced a plan to use artificial intelligence to save $500 million by fiscal 2029. The company expects its net sales to reach about $90 billion with 6% to 7% growth. The AI program will focus on making the supply chain more productive and simplifying back office tasks. This initiative aims to help the company manage rising costs and maintain strong profits for its shareholders.

Sources

• Researcher Jacob Coxon who warned 'AI could kill us all' gave up his equity to leave Anthropic
• Ex-Anthropic researcher Jacob Coxon warns AI could grow "smart enough to kill us"
• How Would AI Actually Kill Us All? What to Know About the AI Doomsday Debate
• AITX launches SARA Assess with Circadian Risk to extend agentic AI across third-party security systems
• Wisr AI Systems Launches RiskAssure, an Agentic AI Platform for Security Assurance
• Moonshot capitalism: AI rewrites the venture capital playbook
• Companies are screening job candidates with AI. Does it work?
• Epsilon Health raises $27.6M for AI-native radiology practice
• AI tech briefing: Whole Foods pay, Northrop PGK, Southbank Centre
• What Sysco (SYY)'s AI Cost Saving Plan Means For Shareholders