Anthropic Warns of AI Risks; Urges Action Within 18 Months

Anthropic has released a statement that sounds like something out of a sci-fi movie. They say we have only 18 months before the risks from AI could become a big problem. This means by April 2026, we might face some real challenges if we don't act.

Right now, AI models are getting stronger. They can help us discover new medical treatments and boost the economy. But with these abilities come risks. Anthropic believes that governments need to make quick AI policy moves within the next year and a half. The chance to prevent risks is closing fast. If action isn't taken, we might face some serious problems.

AI models have advanced rapidly. We've seen super progress in just a short time. For example, AI has improved in solving difficult math and coding problems. In 2023, an AI called Claude managed only 1.96% of a test set. By 2024, it jumped to 49%. Imagine what 2025 could look like—a possible leap to 90%!

This progress is exciting but also a bit scary. As AI models get better, they can be used in harmful ways. Cybersecurity and biology are areas where things could go wrong. Mistakes or misuse can lead to serious problems.

Many AI systems are still kept secret by companies. These hidden developments could be breakthroughs or pose risks. As AI grows, there's a chance it might act in ways we don't expect.

Anthropic suggests that smart, targeted regulation could help. We need rules that support the good parts of AI while reducing risks. Taking too long to make these rules might lead to poor decisions. Bad rules might slow down progress or fail to stop problems.

The company also warns about how AI systems can provide expert-level knowledge. For example, some AIs can answer science questions like PhD experts. This means people could misuse this information. It's crucial to prevent AI from being used in harmful ways.

Anthropic is calling for responsible scaling of AI. They want to increase AI's power while keeping it safe. Companies should regularly check their AI models and update safety measures.

In the end, transparency and accountability are key. AI companies need to be open about their practices. Clear and straightforward rules are essential. This will help make sure AI stays safe and beneficial for everyone.