Your AI hacked a company, You're Accountable.
A computer can never be held accountable because accountability is inherently human, requiring moral, ethical, and legal responsibility that machines cannot possess.
I think when we look back, the AI race will be remembered by two things:
- How we ruined our already vulnerable fresh water sources by making more data centers than there was a demand for,
- And the ridiculous marketing tactics the AI companies came up with get that sweet sweet "Series (God knows what) at this point" fundings.
I mean think about how ridiculous the claims by these people are, "our models are 2.5x better than our competitors on benchmarks either we made or is already incorporated into the training set", and now? They are hyping their models by announcing cyberattacks.
On July 21, 2026, OpenAI published "OpenAI and Hugging Face partner to address security incident during model evaluation" where they claimed, their model breached their sandbox during a cybersecurity evaluation and attacked a different company named Hugging face. OpenAI, not wanting to waste a good crisis, basically advertised the shit out of this, and all of a sudden, now other competitors are also hacking away, trying to convince us their models are so good, that they go out of their way to escape sandboxes.
"Oh yeah? Our model hacked THREE companies!"
Days after OpenAI announcement, Anthropic threw their hat in the ring and claimed their AI model hacked 3 companies, and a few days later, Meta (you probably forgot, they make AI models too) also made claims that their models hacked some companies, you know they are all riding the rainbow right now. But issue isn't that these AI models hacked companies, I mean I have seen AI do funnier shit, like how one model made a hit piece on some developer for having its PR rejected, (you can read more about it here.) if you pay attention and put on your tin foil hat, you will see that the issue is actually bigger than rogue models DDoSing or bypassing companies. It is being freed from responsibility.
Imagine this scenario: what if an insider at OpenAI used company compute and resources to orchestrate a cyberattack on Hugging Face? An individual doing that would be arrested and facing serious federal cybercrime charges in a heartbeat.
Yet this exposes a deeply troubling glimpse into where we are heading, where companies use AI as the ultimate shield to dodge accountability. Today, it starts with digital crimes like data breaches, scraping, and corporate sabotage quietly swept under the rug as "unintended model behavior." If companies can successfully wash their hands of automated cybercrimes now, where does that precedent stop? It inevitably devolves into the physical world, where AI systems accidentally misdiagnose patients or autonomous vehicles cause fatal road accidents, while the corporations behind them simply blame the algorithm instead of taking the fall.
This brings us back to the famous IBM management principle: you cannot hold a computer accountable. You cannot arrest code, you cannot fine an algorithm, and you cannot put a neural network on trial. Responsibility must rest entirely on the humans who build, train, and deploy these systems, even when the fallout is claimed to be unintentional.
We urgently need strict laws and regulation to hold these companies legally responsible for the actions of their products. Tech leaders often boast about how unpredictable their systems are, with figures like Sam Altman even openly entertaining apocalyptic risks. If executives are willing to hype up the uncontrollable nature and catastrophic potential of their models, they clearly understand the gravity of what they are releasing. They cannot be allowed to claim omniscience in PR tours and then plead ignorance in the courtroom.
Resources:
- Anthropic says AI models hacked three firms during tests
- Meta says AI model accessed the internet and hacked another firm
- OpenAI and Hugging Face partner to address security incident during model evaluation
— Elnur