Should AI companies release models powerful enough to carry out cyberattacks, even to vetted customers?
On November 5, OpenAI released GPT-6 Cyber, a model built for cybersecurity work, to a restricted group of approved customers.
On November 5, OpenAI released GPT-6 Cyber, a model built for cybersecurity work, to a restricted group of approved customers.
Votes cast anywhere this poll is embedded count toward the same live total.
Seriously do people not remember that there have been computer hackers for the last 50 years? Jobs and Wozniak paying the bills by selling blue boxes to phone phreakers in the 70s isn't a story of how clever the Steve's were. It's a story about how appallingly bad computer security was then, is now, and forever will be as long as short changing security feeds profits and only hurts people
What a dumb question. All models are powerful enough to carry out cyberattacks. "Should companies be able to release knives sharp enough to cut humans?" "Should gun makers release bullets that can put an eye out?"
Option 3 is revolting; as if governments and "security" aren't prime suspects for who'd actually abuse it.