Yoshua Bengio

I was pleased to give remarks at the start of the Digital Summit in Berlin this afternoon. Frontier AI companies continue to push model capabilities further and faster forward, without a convincing plan to make the technology safer. This dynamic carries significant risks to collective safety, including malicious misuse, loss of control and other misaligned behaviors. Alignment and safety must be treated as required capabilities to protect the public. We urgently need strict and independently validated red lines, research into safe-by-design AI, such as the work we are doing at @LawZero_, and urgent international cooperation - a path Germany and Canada are actively charting together.