נושאים
נושאים
החוטים שעוברים בכל מה שנכתב כאן — ולכל אחד מהם כל הדפים שעוסקים בו.
EU Machinery Regulation
The EU regulation that replaced the Machinery Directive and, for the first time, puts software, autonomy and self-evolving behaviour inside the scope of a machine's safety assessment. It applies from January 2027, and it is the instrument that decides which machines a third party must inspect before they may be sold. Everything written here about conformity routes, substantial modification and inspection capacity traces back to it.
דפים שעוסקים בזה · 43
Cybersecurity for physical AI
When a machine can move, a security flaw becomes a safety hazard: an attacker who reaches the controller reaches the actuators. Two EU instruments now carry obligations that outlive the sale — a support period for security updates, and incident reporting duties for operators of critical systems. This hub collects what is written here about where safety engineering and security engineering stop being separate disciplines.
דפים שעוסקים בזה · 16
Functional safety
The discipline of making a system behave safely when a part of it fails — defined by its standards, measured by its integrity levels, and argued for in a safety case. The glossary entry defines the term; this hub gathers every page that works with it.
דפים שעוסקים בזה · 63
Safety standards
A regulation states a requirement; a standard states how to meet it, and a harmonised standard carries a presumption of conformity in European law. The gap between the two is where most of the practical difficulty of Physical AI safety lives, because the standards that exist were written for machines whose behaviour was fixed at the factory.
דפים שעוסקים בזה · 51
Certification
The act of demonstrating to someone other than yourself that a machine is safe enough to sell. For Physical AI this is unusually hard: the evidence a third party is equipped to inspect was designed for behaviour that does not change after shipping. This hub collects the routes, the capacity constraints and the open questions.
דפים שעוסקים בזה · 35
EU AI Act
The EU's horizontal law on artificial intelligence, which classifies systems by risk and attaches obligations to each class. Its relationship to machine safety law is the part that matters here: a robot can be in scope of both, and the two frameworks ask for different evidence about the same behaviour.
דפים שעוסקים בזה · 6
The physical AI industry
Who builds physical AI, where, and with whose money — the national and regional ecosystems, the venture funding behind them, and the safety-infrastructure category no funding tracker counts.
דפים שעוסקים בזה · 10
Incidents and liability
What happens after a robot hurts someone — the incident record and the pattern in it, how liability splits along the chain from AI vendor to operator, and the insurance models that have not caught up.
דפים שעוסקים בזה · 10