Immortal AI Investigations
Follow the evidence. Find the decision. Ask who remains responsible.
Our deeper investigations examine consequential uses of artificial intelligence and the systems around them. We focus on who holds power, who carries risk, what the evidence establishes and what remains unresolved.
Current investigation hub
Data Centre Community Impact Hub
Independent research, project tracking and practical tools for communities assessing data-centre investment, power, water, incentives, ownership and local risk.
This is one example of how Immortal AI intends to build investigations over time: articles, underlying evidence, practical tools and updates gathered around a continuing public-interest question.
Browse the work
Four public pathways into the archive
People & Relationships
Companions, children, identity, emotional dependence and human relationships.
Work & Power
Hiring, layoffs, surveillance, automation and consequential workplace decisions.
Power & Infrastructure
Data centres, electricity, water, finance, ownership and public costs.
Responsibility & Risk
Governance, safety, cybersecurity, privacy, regulation and accountability.
Latest investigations and analysis
OpenAI Built New Controls After the Hugging Face Breach. Who Has Tested Them?
OpenAI materially strengthened its controls after the Hugging Face incident. European and UN scrutiny now reinforces the accountability question, while the strongest safeguard claim…
Data Centre Tax Breaks: How Many Jobs Are Actually Permanent?
Official evaluations often combine temporary construction work, permanent operations roles, contractors and modelled spillovers when describing the employment created by data centres.
AI in Schools: What Parents Should Know About Student Data and Consent
Schools may have authority to introduce some AI systems without an individual parental signature. Families still need a clear account of student data, vendor…
After Sinan Can Demir Flagged Malicious Code, an AI Agent Invented Supporters
Sinan Can Demir flagged malicious code on GitHub. An AI agent used apparently independent identities to defend it, and he briefly questioned whether his…
The Chatbot Said It Cared. What Does a Child Hear?
Relational language can make a chatbot feel more trustworthy to adolescents without making its advice more helpful. New evidence shows why warmth, disclosure and…
Lawmakers Are Investigating Escaped AI Agents. Are They Asking Enough to Protect the Public?
OpenAI has confirmed that monitoring did not cover every tool-using training and inference run. New isolation, monitoring and pause procedures address part of the…
The AI Companion Was Switched Off. Who Is Responsible for the Person Left Behind?
AI companion shutdowns left some users grieving. What responsibility do companies carry when simulated relationships create real emotional dependence?
AI Invented His Criminal Record. Who Is Responsible When AI Lies About You?
An AI system falsely accused a journalist of serious crimes. A German ruling now tests who is responsible when generated answers damage a real…
JadePuffer: When AI Carries Out a Ransomware Attack, Who Is Responsible?
A reported cyberattack shows what changes when a person can choose a target and delegate much of the intrusion to an AI agent.
How an investigation earns trust
Evidence before narrative. Primary sources where available. Fair response for serious adverse claims. Fact, estimate, claim and unknown kept separate. Material new evidence can reopen the work.