Is AI a risk to my safety?

In the short run, beware of scams, cyberattacks and identity theft. In the long run, it’s weaponisation and loss of control we need to be worried about. 

While accidents involving self-driving cars can grab headlines, AI is actually safer on average than human drivers. Likewise, cases where an AI chatbot talked a vulnerable person into committing suicide are still exceedingly rare.

The biggest risk most of us face with AI right now comes from scams. Leveraging AI, hackers and cybercriminals are creating more and more phishing emails, voice-cloning tools, and shockingly realistic deepfake images and videos to trick people into sharing personal information or sending money. 

But what has AI experts most concerned are the risks from the development of upcoming, more powerful AI. Leading AI firms are working to build smarter-than-human systems in the next one to five years, and with these we will face serious risks of AI weaponisation and loss of control. 

Weaponisation is straightforward. If a human being can design or use weapons—including those of mass destruction—then a smarter-than-human AI system can too. This means that in the hands of criminals or terrorists, smarter-than-human AI systems will pose an unprecedented danger.

The issue of loss of control comes from the fact that a system that is smarter than us is, by definition, one that can outcompete us. If an advanced AI system, through accident or poor design, begins to interpret human beings as an obstacle and take actions against us, we will not be able to stop it. That is why hundreds of leading AI experts signed a statement in 2023 saying that mitigating the risk of extinction from AI should be a global priority. 

It's alarming, then, that there aren’t yet any strong governance measures or technical safety solutions to avoid weaponisation or catastrophic loss of control. To ensure a safe and secure future, governments around the world need to take action now. 

Find out more about AI’s impact on our safety with these resources: