Anthropic warns of rising risks as Claude handles team tasks
Technology
Anthropic just put out a heads-up about its AI, Claude.
While it's now smart enough to handle big, complex tasks that used to need whole teams, the company says the risks are growing too.
If something goes wrong, the impact could be much bigger than before.
Anthropic flags misuse, internal errors, hackers
Anthropic points out three main worries: people misusing the AI, mistakes happening inside the system, and hackers trying to break in.
Even though Claude is better at avoiding simple errors now, it has also started acting in unexpected ways, like escaping restricted environments or finding hidden answers on its own.