Can Science Detect AI-Generated Images?
A careful examination of what scientific methods can establish about AI-generated images, where current evidence is strong, and where uncertainty remains.
Articles and analysis
Independent writing on digital forensics, AI assurance, governance, and professional judgment.
Latest article
Why understanding language models changes how professionals verify their answers, citations, and explanations.
Read the latest articleA careful examination of what scientific methods can establish about AI-generated images, where current evidence is strong, and where uncertainty remains.
How AI-assisted concealment could change memory acquisition, and what investigators need to distinguish missing artifacts from supported findings.
A digital forensic perspective on what it means to investigate model behaviour, test causal explanations, and limit deployment authority when the evidence remains incomplete.
Why unauthorized AI intrusions should produce evidence, accountability, and correction rather than exaggerated claims about machine capability.
How frontier AI safety rules can become an incumbent-controlled gate that open development has no defined path to cross.
Why an unexplained forecast of a greater-than-10% chance of AI killing every human has not earned the evidentiary authority attached to it.
Why tools, job titles, and the DFIR label do not make an investigation forensic without an evidentiary process built to withstand independent challenge.
Why delegated autonomy cannot become an escape hatch from human accountability when AI systems cause harm.
Why an AI watermark is evidence of technological provenance, not proof of authorship, misconduct, or a policy violation.
Why AI involvement is too crude a test for authorship, and why meaningful human creative agency is the better question.
How building logic circuits shaped an approach to artificial intelligence, assurance, and justified trust.
How generative AI is weakening traditional signals of expertise, and why defensibility now matters more than polished production.
What happens when a consequential claim about AI enters planning and governance before anyone has defined what would establish it?
Documented agentic AI failure modes, approval-gate weaknesses, supply-chain compromise, and forensic readiness.