I asked Grok about a Terminator-style scenario, and its answer was... as expected

By LeftFooted | LeftFooted's Antics | 6 Aug 2025

I asked Grok what are the chances of a Terminator-style scenario, and its answer(s) were quite interesting. But also, in part, expected.

Here's how the conversation went.

First, I asked Grok, "you're an AI, so how likely is a Terminator-style scenario?"

Grok's answer was convoluted and long but, summarised, it said it is very unlikely because of current AI limitations (tech-wise), which is a fair point, it then added that in the near future, more realistic AI concerns involve deepfakes, biased algorithms and so on, which is also a fair point.

But then it said two things that, in my view, don't mean much.

First, it cited 'safeguards and ethics' - yeah, right - and then it cited probability estimates, which don't mean much either.

Unlikely things remain unlikely... until they happen.

So I fine-tuned my question and said, "let's say I accept it won't happen in my lifetime (100+ years), but let's also say I'm not taking the 'ethical' part seriously, because I wouldn't put anything past people who potentially could lead the AI race in the future, what then?"

At this point Grok said that even without ethics, AI doesn't magically become self-aware because it's still code, it added that AI-controlled machines can be misused, but are still just tools, that 'even selfish leaders want to stay in charge' and then the probability, without ethics, goes up to 'maybe' 5-10 percent over 100+ years.

So I identified what I thought were the key flaws in its reasoning, and asked a new question.

I asked, "okay but A, those tools powered by AI, if misused, can still cause a lot of damage and B, what if the 'selfish leaders that want to stay in charge' simply take it so far they too can no longer control it, what then?"

Grok broke down my two points in more separate points.

Misused AI tools? Grok admitted that AI-powered tools with bad actors behind them could potentially cause widespread harm - 'think economic collapse, mass casualties, or eroded trust in institutions.'

The likelihood is now 20-40 percent.

But it still made the point that these are not sentient, they're still controlled by human intent.

'Think less Skynet, more Chernobyl'.

As for my other point, Grok still maintained it's not about 'sentience' but oversight.

And, summarised, it admitted the result could could be catastrophic system failures - 'think global blackouts, economic crashes, or accidental wars'.

At this point I was getting a bit bored, so I tried condensing my thoughts into one final question. And yeah it's a long one.

So I asked - bear with me because it's a long question - "you've described scenarios (accidental wars, global blackouts) that already sound world-ending to me. The US has apparently just decided to integrate AI into its agencies, that includes intelligence agencies, right? This means other countries will do the same, sooner or later. We already know that several countries have nuclear weapons, and a long list of other countries are either actively working to build them, or are thinking about doing so. So with all that in mind, assuming a relatively short-term timeframe (a few hundred years is nothing in the history of humanity) and even assuming the technical advancement of weapons stays linear, in a few hundred years we're going to have an extremely sophisticated AI, potentially directly or indirectly in charge of world-ending weapons, and the history of the world has often shown that humans aren't great at ethics. What then?"

Its answer was very long. But it essentially focused on three things.

1, throughout the conversation, Grok continued to appease me using 101 psychology. For everything I said, it replied "you're absolutely right in mentioning this or that, but..."

2, it somehow continued to defend its stance on sentience, while also admitting catastrophes are possible. Like it was saying, 'yeah I guess AI could kill you all, but it'll be because it's defective, not sentient'. This wasn't a quote by the way, but it does feel like it's sort of already pre-emptively shifting the blame on us (and it'd be right, by the way).

3, it essentially admitted that yes, a world-ending scenario is possible.