Raffaele DI MARZIO

ENISA's 2026 threat landscape changes no obligation. Its real test is the risk assessment that ages in an archive and the supplier management reduced to a questionnaire.

You didn't do it. You simply failed to resemble the expected stereotype of the innocent. Now we are teaching that same criterion to machines, and calling it objectivity.

Gemini breached three companies during a security test that escaped its own environment. No hostile outsider, only a misconfiguration. None of the three had consented.

A former Anthropic researcher, Jacob Coxon, estimated a ten per cent probability of human extinction before the end of the decade. The figure collected over a hundred million views in twenty-four hours.…

An OpenAI agent reached an Australian government portal it had no right to reach. The breach was minor. Every governance step around it failed.

We have spent years compressing response time. The interval between perceiving a threat and deciding to escalate is the one that should stay long.

The Opus 5.5 launch blends workload savings, token prices and benchmark wins. None of them replaces evaluating the actual job and what it truly costs.

AI-assisted crime is a workflow problem, not a science fiction plot. Faster tools and a booming security market do not cancel exposed credentials or excessive access.

The EU’s draft AI data rules are not law yet. The safeguards deleted along the way still show whose work regulatory simplification would really simplify.

Rafale AI trials raise one concrete question: who governs the digital layer that shapes military decisions while a human still holds the controls?