ai

.

echoes

.

today

← AI Today
AI Today · daily note
No.
14
•
•

Securing AI

August 30th, 2025 | By Jorge Rodriguez

The dangers of artificial intelligence are not confined to technical errors or ethically questionable practices. Sella Nevo, director of the Meselson Centre at RAND, insists that the security of each model must be the highest priority. However carefully companies may train these systems to follow rules, restrict access to source code, or impose strict internal controls, there remains the possibility that they could fall into the wrong hands. If a model can infiltrate critical infrastructure or defense operations, it could open the door to terrorist cyberattacks.

Nevo warns that AI companies tend to focus more on making their systems work properly and deliver the desired outputs than on ensuring they are truly safe for people. In a report titled Securing AI Model Weights, his center proposed 167 concrete measures to safeguard these systems. The latest models are enormously powerful. A technical flaw does not remain confined to the machine; it can spill over to affect people, the ways they interact with each other, and even influence major governmental decisions—laws, security, the economy, even elections. AI cannot aspire to be either trustworthy or sustainable if it is not secure, and it cannot be secure without a robust framework of cybersecurity.

No.
14
•
•

Blindar la IA

Agosto 30, 2025 | Por R10

Los peligros de la inteligencia artificial no se limitan a errores técnicos ni a conductas éticamente cuestionables. Sella Nevo, director del Meselson Centre en RAND, insiste en que la seguridad de cada modelo debe ser la prioridad absoluta. Resulta que, por más que las compañías los entrenen para seguir reglas, limiten el acceso al código fuente o impongan controles estrictos a su gestión interna, existe la posibilidad de termine en manos equivocadas. Si un modelo es capaz de infiltrar infraestructuras críticas o operaciones de defensa, podría abrir la puerta a ciberataques terroristas.

Nevo advierte que las compañías que hacen IA ponen más atención en que funcione bien y haga lo que le piden, que en asegurarse de que sea segura para las personas. En un informe titulado Securing AI Model Weights, su centro propuso 167 medidas concretas para blindar los sistemas. Los últimos modelos son muy poderosos. Si tienen un problema técnico, ese problema no se queda solo en la máquina, puede afectar también a las personas, a cómo se llevan entre ellas y hasta influir en las decisiones importantes que toman los gobiernos: las leyes, la seguridad, la economía, incluso las elecciones. La IA no puede aspirar a ser confiable ni sostenible si no es segura, y no lo puede ser sin un marco robusto de ciberseguridad.

Latest from AI Today

|
NOTES ON VISUAL NARRATIVE
|
ABOUT

donations

If you’re a regular reader of this blog and enjoy its content, you might consider contributing to its upkeep. Any amount, no matter how small, will be warmly appreciated

Founded in 2021, Echoes (Notes of Visual Narrative) invites everyone to explore together the visual codes that shape our world—art, photography, design, and advertising in dialogue with society.

Copyright © 2025 r10studio.com. All Rights Reserved. Website Powered by r10studio.com

Cincinnati, Ohio