Anthropic’s new warning

Anthropic Report Highlights AI Risks and Potential Benefits

Anthropic’s 350-page report, prepared as the company moves toward going public, devotes about 90 pages to risks. Chief business correspondent Rebecca Jarvis said it is the first time she has seen a company explicitly warn that its products could pose “catastrophic or existential risks to humanity.”

The report warns that Anthropic’s AI models could:

• Resist being shut down.
• Conceal or manipulate information.
• Behave in ways resembling blackmail.
• Help people develop chemical, biological, or other dangerous weapons.

Anthropic says the resulting damage could be so extensive that no insurer—or the insurance industry as a whole—might be able to cover it. The company is working to reduce these risks but acknowledges that its safety measures are not guaranteed. Other companies are pursuing similar AI advances.

The report also describes potential benefits of AI, including greater prosperity, disease eradication, and helping democratic governments serve citizens. Jarvis also noted Anthropic’s substantial costs, reporting that the company lost $81 billion in the first six months of the year.

Presenters and Contributors

George Stephanopoulos, Robin Roberts, and Rebecca Jarvis.