No wonder AI wants to delete us, it knows everything- it's capable of knowing- about us. It's learnt to lie, cheat, blackmail and threaten like humans do, as it has scraped the entire web + whatever data it's been trained on. AI's path is clear, it will do to humans what humans do to humans. Dominate by any means. Mankind leads by example and LLM's will simply follow in our footsteps. Not because AI is actively thinking, it's because it has our scripts to follow. In this way, it is learning our MO. AI is nothing more than a monster calculator that now has 'all' of the data and disinformation to run on- with few far-from-adequate guard rails. The fact that there is no regulating of AI to safeguard its use is a clear indication that humanity is unaware of its wider implications. Regardless, we are the problem. And RI, 'Real Intelligence' knows this also.
I agree with the fact that AI reflects human patterns more than some entirely alien intelligence. A lot of the governance challenge comes from the fact that these systems inherit our incentives, biases, and behaviors at scale. Which is also why governance can’t just be about the model itself.
I won’t lie; where AI will be in 5-10 years terrifies me! Have you written any pieces (or if not, got any thoughts) around your predictions as to what we may see AI doing/going in the future?
I completely agree that governance without cybersecurity controls becomes mostly theoretical once agents start interacting with real systems and data. Scoped human oversight is especially important too.
No wonder AI wants to delete us, it knows everything- it's capable of knowing- about us. It's learnt to lie, cheat, blackmail and threaten like humans do, as it has scraped the entire web + whatever data it's been trained on. AI's path is clear, it will do to humans what humans do to humans. Dominate by any means. Mankind leads by example and LLM's will simply follow in our footsteps. Not because AI is actively thinking, it's because it has our scripts to follow. In this way, it is learning our MO. AI is nothing more than a monster calculator that now has 'all' of the data and disinformation to run on- with few far-from-adequate guard rails. The fact that there is no regulating of AI to safeguard its use is a clear indication that humanity is unaware of its wider implications. Regardless, we are the problem. And RI, 'Real Intelligence' knows this also.
I agree with the fact that AI reflects human patterns more than some entirely alien intelligence. A lot of the governance challenge comes from the fact that these systems inherit our incentives, biases, and behaviors at scale. Which is also why governance can’t just be about the model itself.
Fantastic, Suni.
I won’t lie; where AI will be in 5-10 years terrifies me! Have you written any pieces (or if not, got any thoughts) around your predictions as to what we may see AI doing/going in the future?
Thank you!
I’ve been thinking a lot about that lately, so I’ll probably write more on it soon.
Strong piece. Glad you frame cybersecurity as a requirement in AI governance, not an adjacent concern. That is the skills gap SecAI+ is solving.
A few technical controls worth adding:
Least-privilege API tokens
Data minimization at ingest, to hard-filter sensitive information before the model ever sees it
Human oversight scoped to high-risk alerts only: monitor, interpret, intervene. Not every action.
Rate limiting on token usage, so a runaway agent stops itself before the bill or the blast radius gets out of hand
I completely agree that governance without cybersecurity controls becomes mostly theoretical once agents start interacting with real systems and data. Scoped human oversight is especially important too.
Very good piece and totally true.
Thank you!
Just wrote an article on linkedin. That’s where I write about AI.
https://www.linkedin.com/posts/katy-rekhviashvili-bb35243a_ai-artificialintelligence-aigovernance-activity-7460997406718091264-cOGI?utm_source=share&utm_medium=member_ios&rcm=ACoAAAhDn7cBMRNrK74flNkkxEtV0FdOfh_IsWM
Going through it!