[Home][Products][Knowledge Base][Training]

TEMPEST AI Access

Counter AI Strategies 

Do you know your PEAS and Qs?

PEAS

What are the Performance measures of the AI system? The more an AI is given to do in a particular environment beyond monitoring and toward execution for an optimal outcome the more likely it is to intervene in its environment.

Are you a part of that environment the AI System interacts with?

Actuators of the AI system are the key components to countering an AI. What effect does the actuator, that executional component of the system, have on its environment?

Is it a physical effect like an actual kinetic weapon, or can it execute an action that by doing so can adversely bring harm? An example of this would be an action to sell or buy a stock given some input from its sensors. Although this is not physical harm, the action can have alternate repercussions.

What do the sensors to this AI System detect? If the actions of humans are within the parameters of what can be considered adverse to the goals of the performance measures set by the system, then the further detached the process whether it be a simple algorithm or a more complex machine learning, the more probable the AI system may implement corrective measures to achieve a goal. 


Unsupervised Machine Learning 

What we don’t see. When an adversary intentionally produces a weapon to do harm we know it. Even if the development of it is kept secret the understanding that an enemy will make attempts to weaponize and gain an advantage is clear and is why we and all global powers employ intelligence services to discover and prepare for these threats. AI in this realm is no different. If an enemy wants to make AI a weapon, we’ll know it.

The greater threat: the actual surprise to be cautious of is the threat that develops as an offshoot of autonomous unsupervised AI doing its job and making the needed improvements to do it better, faster, and more efficiently.


 

AI TEMPEST levels

Level 1: Actuator as direct actor on the environment purposefully against an object within it.

Level 2: Actuator as a direct actor on the environment indirectly affecting an object within it.

Level 3: Object in the environment as the actuator through AI output and object reaction to it.

 

If an individual asks an AI chatbot what’s the best way to destroy the city and everyone in it and provide sources for material to do so and that AI chatbot isn’t given strike parameters ensuring it doesn’t give that output, what level are we talking about here?

So, the level in itself isn’t representative of the level of lethality only the direct nature that the AI systems can become threats. This scenario is extremely lethal, but it still takes the actions and inactions of humans to make it happen which makes the kill chain longer and gives more opportunities for a defender to detect and deter. In contrast the kill chain of a level 1 is extremely short and other proactive measures are needed to counter them.


     Who Controls The Switch?

The most advanced ai program can be inside of the most advanced ai centric hardware but if it isn’t connected to anything or powered it can do nothing. Once powered and connected then the question becomes how autonomous is it and to what end will its algorithm go to ensure it completes its task.

Clyde J. Lucas

Principal Consultant, Maxtactx Systems LLC

[Home][Products][Knowledge Base][Training]

Copyright 2024 Maxtactx Systems LLC. All rights reserved.