The next time you visit your therapist, he or she may ask you a simple question: How is your relationship with the AI model?
If you say it’s complicated, you might have a point.
Let’s start with cybersecurity. On the digital side, geopolitical cyberattacks have rapidly increased in the wake of the Iran war, and attacks are now being deployed in parallel with physical weapons.
And with the rise of AI deepfakes and highly personalized phishing emails, experts warn that we can no longer rely on what we see and hear.
So what about strategy simulations? Before going to the battlefield, the AI is tested in war games.
In a recent study, models including ChatGPT, Claude, and Gemini were exposed to simulated crises.
And the results were alarming, with at least one AI in every game escalating the conflict by threatening to use nuclear weapons.
Finally, in the recent Iran airstrike, the U.S. military relied on Anthropic’s Claude to identify targets.
But when Anthropic denied unrestricted access to the Department of Defense, citing ethical concerns, OpenAI quickly stepped in and awarded the contract.
OpenAI argues that the agreement strictly prohibits mass domestic surveillance and requires human oversight of weapons.
OpenAI’s Sam Altman defended the deal, writing, “We remain committed to serving all humanity to the fullest. The world is a complex, messy, and sometimes dangerous place.”
Therefore, this agreement is likely to result in either greater involvement with humanity or a more complex and dangerous situation.
And as I said earlier, it’s “complicated.” Right, ChatGPT?
For the full story, watch the Euronews video in the player above.
