A New Trick Could Block the Misuse of Open Source AI
Briefly

Terrorists and rogue states are going to use these models. The easier it is for them to repurpose them, the greater the risk. - Mantas Mazeika, Center for AI Safety researcher
Powerful AI models are often kept hidden but releasing them opens risks. Meta released Llama with parameters available, necessitating safeguards. - Wired
Read at WIRED
[
|
]