Anthropic Researcher Warns: AI Could Kill All Humans by 2030

Key Takeaways

• Jacob Coxon resigned from Anthropic after warning that AI could kill all humans by the end of the decade.
• Coxon's resignation post received over 160 million views and generated 18,000 comments from peers.
• Anthropic safety lead Evan Hubinger estimates a greater than 10% chance of AI extinction within the next decade.
• OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei warn that AI development must slow down for societal preparation.
• Elon Musk labeled Coxon's warning a marketing stunt, though the message reached over 120 million views.
• A ByteDance study found only 34 of 64 code changes made by LLMs worked correctly, including tests on Gemini 3.1 Pro.
• Many small businesses stop using AI tools within the first 30 days because they fail to solve root problems.
• Europe faces dual compliance requirements under the GDPR and the new EU AI Act for high-risk systems.
• Singapore is considering joining a China-led AI governance group, potentially affecting US supply chain partnerships.
• Security leaders should measure AI performance using coverage and accuracy metrics rather than just speed.

Charlotte residents worry about AI safety after researcher quits

Charlotteans are feeling cautious about artificial intelligence after Jacob Coxon, a former AI researcher, resigned from Anthropic. Coxon posted on X that both OpenAI and Anthropic are racing toward self-improving superintelligence and could kill everyone by the end of the decade. Local residents like Dusten Pulido and Jhaya Dilworth agree that safety is not discussed enough as data centers grow. Big tech leaders including OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei also warn that AI development must slow down to allow society to prepare for these new risks.

Elon Musk calls ex-Anthropic researcher warning a public relations stunt

Elon Musk mocked Jacob Coxon, a former Anthropic researcher who warned that AI could kill humans by the end of the decade. Musk suggested Coxon's post was a marketing stunt or a conspiracy, even though the message received over 120 million views. Coxon responded to Musk on X, stating his resignation was for ethical reasons and not a publicity move. Other AI researchers support Coxon's concerns, with Anthropic safety lead Evan Hubinger estimating a greater than 10% chance of AI killing all humans within the next decade.

AI resignation sparks global debate on potential human extinction

The resignation of 27-year-old Jacob Coxon from Anthropic has pushed the debate about an AI apocalypse to new heights. Coxon stated that neither OpenAI nor Anthropic is acting responsibly and that the technology could kill all humans by the end of the decade. His post on X was viewed more than 160 million times and generated over 18,000 comments from peers who supported his claims. Anthropic alignment science lead Evan Hubinger agreed, saying the risk of AI killing all humans is more than 10% within the next decade.

Most AI experts agree researcher warning about human danger is valid

Jacob Coxon, a 27-year-old AI researcher, resigned from Anthropic after posting that the technology could kill all humans by the end of the decade. His social media thread received over 100 million views in less than 24 hours, and many in the field echoed his warning instead of dismissing him. Coxon explained that he spent three years at OpenAI and Anthropic and believes executives fear the risks privately. Anthropic spokespersons admitted the technology brings unprecedented risks, while OpenAI officer Chris LeHane argued that waiting too long to take action is the greater danger.

Security leaders must track more than speed to measure AI SOC success

Security teams might miss threats if they only judge AI performance by how fast it works. Low escalation rates can hide false dismissals and create operational risks for organizations. Security leaders should measure AI Security Operations Centre performance using metrics like coverage, accuracy, and average time to containment. These specific numbers help ensure the AI system performs as intended and allows teams to fine-tune the software until it is optimal.

Ferris State Tech Week explores AI in health and rural areas

Ferris State University is hosting Tech Week Grand Rapids to discuss how artificial intelligence changes health care and society. The event includes about 160 free sessions, workshops, and networking mixers open to the public. Specific talks will cover closing the AI divide in rural and Tribal communities and strategies for leading successful AI adoption in businesses. Another session will share case studies on using AI to engineer polyketide synthases for making medicines and fuels. These events aim to show how technology transforms patient care while addressing gaps between urban and rural communities.

Klook founder says AI helps solve overtourism concentration issues

Ethan Lin, co-founder of the travel platform Klook, stated that overtourism is a concentration problem rather than a volume problem. He argued that too many tourists visit the same few places instead of exploring a destination fully. Lin believes AI can help travelers find hidden gems and new experiences to spread crowds out. The Fortune Leaders Forum in Macau also discussed how influencers promote specific resorts like Myoko Suginohara Ski Resort in Japan. Travelers today often seek unique experiences like celebrity concerts or sports events, which drives demand for flexible travel packages.

Many small businesses stop using AI tools after 30 days

A commentary explains why many small businesses abandon AI tools within the first month of purchase. Companies often buy a tool and receive a quick tutorial, but they stop using it by day 30. The reason is that the tool was never the real solution to their specific problem. Instead of fixing the root issue, the software acted only as a temporary Band-Aid fix. This part of the AI story often does not appear in official case studies because the tool failed to deliver lasting results.

Singapore considers joining China-led AI group amid US tensions

Singapore is weighing an invitation to join the World Artificial Intelligence Cooperation Organisation, a group led by China. The Shanghai-based bloc focuses on AI governance and technical standards. Singapore's Minister for Digital Development and Information, Josephine Teo, said the decision depends on the group's objectives and how it fits with existing partnerships. This move could affect Singapore's role in the US-led Pax Silica initiative, which aims to secure supply chains for AI minerals. Analysts note that joining the Chinese group might bring influence but could also risk US retaliation.

Europe must balance AI innovation with GDPR and EU AI Act rules

Companies in Europe are adopting agentic AI quickly, but governance struggles to keep up with the speed. Gartner predicts that regulatory violations will cause a rise in tech-related lawsuits soon. Only 23% of GRC leaders feel confident managing these risks as autonomous AI tools take on more complex tasks. Organizations must now follow two strict rules: the GDPR for data protection and the new EU AI Act for high-risk systems. Compliance requires handling automated decision-making, impact assessments, and traceability under both frameworks at the same time. A company called Hyperproof offers solutions to help businesses meet these dual obligations while continuing to innovate.

Global Privacy Laws Strengthen Enforcement in 2026

Data privacy laws are becoming more powerful and active around the world in 2026. Regulators are now enforcing rules at a large scale instead of just writing them. The European Union's General Data Protection Regulation has been in effect for ten years and remains a major standard. New laws in places like Colorado address challenges created by artificial intelligence systems. Officials are also looking closely at how companies ask for user consent to fix unfair practices.

ByteDance Study Shows LLMs Struggle to Build Reliable Agents

A new project called HarnessDev tested whether large language models can build their own agent harnesses. Researchers from ByteDance Seed and other universities found that only 34 of 64 code changes made by the models worked correctly. The study tested six different models including Opus 4.8, GPT-5.5, and Gemini 3.1 Pro across four domains. Many of the code pieces created by the models were inactive or did not function during real tests. The results suggest that current AI models cannot yet reliably engineer their own tools for complex tasks.

Opinion Column Urges Pause on AI Development

Peggy Noonan wrote an opinion piece arguing that humanity should pause artificial intelligence development. She believes the technology is moving too fast and carries dangers that are very close to reality. The author compares the current situation to climate alarmism and says the topic is more important than anything else today. Noonan suggests that people need to understand the critical moment they are facing before proceeding further.

Sources

• Charlotte neighbors react to former AI researcher’s concerns the technology could cause serious harm
• Musk Touts ‘Psy Op’ And Mocks Ex-Anthropic Staffer Who Warned AI Could ‘Kill Us All’
• AI: Incidents and Anthropic researcher's resignation push 'AIpocalypse' debate to new heights
• Few dispute researcher's warning that AI poses threat to human life
• How security leaders should measure AI SOC performance
• Ferris State’s Tech Week Grand Rapids sessions focus on the growing role of artificial intelligence
• AI solves overtourism’s ‘concentration problem’ by helping people find hidden gems, says Klook cofounder Ethan Lin
• The AI tool you will stop opening by day 30 isn’t the answer
• As Singapore weighs joining China-led AI body, does it risk US retaliation?
• Balancing AI Innovation with GDPR and EU AI Act Guardrails in Europe
• As AI Expands, Privacy Law Struggles to Keep Pace
• Can LLMs Engineer Their Own Agent Harness? ByteDance Seed’s HarnessDev Says Only 34 of 64 Changes Generalize
• Opinion | Pause AI for Humanity’s Sake