Reliability & Safety
With emergence of Machine-Learning systems and sociotechnical organisations, it becomes even more necessary to break away from the assumption that systems security is a post-incident-response or has little societal consequences.
This would require instituting a learning culture which befits the complex interdependence between people and the increasingly autonomous systems. Towards this goal, the central ideas in our toolbox constitute the development and advocacy of:
- Data Realism
- Energy Efficiency
- Technology Governance
"People, Ideas, Machines... In that order." - John Boyd
The above quote directs our holistic approach to use an internally designed STAMP based framework in building safety into emerging human-machine environments.
What We Do
Towards greater data realism and energy efficiency, our methodology has been designed specifically keeping in mind the need to handle non-uniformity of applications across the spectrum of AI deployment environments. Given the limited interpretability of neural networks and the dynamic risk environment, our functions can be broken into following three verticals:
Research
Rooted in system architecting and risk management principles, we research and carry out a detailed nonpartisan security analysis to separate the adversarial variables and the technical reality that can inform public policy as well as the secure design of future autonomous systems.
Audit
Our unique Machine Autonomy Risk Mitigation Approach investigates all layers of AI operations; whether physical, social, network, firmware, application, or the environmental layer of its actuation. A probabilistic risk audit across all layers ensures hierarchical safety and continued reliability.
Advise
Assessing the security implications of autonomous agents involves advising about risks at both macro and micro levels. These may have associated strategic, operational, financial, governance, or geopolitical factors. Hence we are more like a think-tank than a conventional advisory.
Contact Us
Do not hesitate to reach out to us.
Fill the form below. Or email us [ hello@asatae.foundation ] for a quicker response.