During the Fortune Brainstorm Tech conference, Dario Amodei, CEO of Anthropic and former OpenAI employee, tackled the concerns surrounding AI risks and outlined a comprehensive categorisation.

Highlighting short-term, medium-term, and long-term risks, he shed light on several concerns that he categorised into three tiers. Short-term risks include issues like bias and misinformation. Medium-term risks arise as AI models advance in fields like science and engineering, potentially enabling harmful applications. Long-term risks emerge when models gain agency and autonomy, making it difficult to control their actions, leading to existential threats.

Amodei emphasised the importance of designing AI systems with safety in mind to mitigate negative consequences. While he expressed optimism about the future, he acknowledged a 10% to 20% risk of things going wrong, underscoring the need for proactive measures.