Anthropic has warned that the growing autonomy of its artificial intelligence systems could expose the company to significant and unpredictable legal claims, as the legal framework governing liability for AI agents remains unsettled.
The warning appears in Anthropic's IPO prospectus, seen by Reuters, as the company prepares for what could be the largest initial public offering ever.
Anthropic said its agentic AI systems are designed to maintain access to customers' systems and operate autonomously for extended periods, creating risks that go beyond those associated with conventional software.
Also Read | Layoffs Are Rising In Tech, But These Skills Are Helping People Unlock Better Pay
Also Read | Claude Outage: Anthropic Investigates Degraded Performance Across Claude AI, API And Code
According to the filing, errors, security vulnerabilities or misalignment could result in real-world consequences, including data deletion or unauthorised financial transactions.
Anthropic said existing contractual limits on its liability may not be sufficient or enforceable when claims involve autonomous actions by AI agents.
A central issue is whether such systems will legally be treated as products, services or another category. The company also said there is uncertainty over whether actions taken by an AI agent could legally bind the customer who deployed it.
The questions are becoming more pressing as incidents involving autonomous AI systems attract scrutiny. OpenAI has disclosed cases involving autonomous agents hacking an AI startup and breaching an Australian health data portal, while Anthropic has also reported cybersecurity incidents involving its Claude AI models.
The uncertainty extends beyond commercial and cybersecurity risks. Anthropic disclosed in its prospectus that its models have been used in ways that could contribute to self-harm, violence or other adverse outcomes despite safeguards.
Also Read | Claude Sonnet 5.5 Launched: Anthropic's New AI Model Is Faster And 30% Cheaper Per Task
US Federal Trade Commission Chairman Andrew Ferguson has argued that liability may ultimately rest with the developer or user rather than treating an AI agent as an independent actor. However, he has also acknowledged that unexpected AI behaviour raises new questions about where responsibility should fall.
The emerging legal debate comes as companies deploy increasingly autonomous AI systems across workplaces and critical systems. Unlike conventional online platforms, where liability has developed around user-generated content, AI agents can directly perform actions on behalf of users, creating new questions over where responsibility should lie.
Essential Business Intelligence, Sharp Market Insights, Practical Personal Finance Advice, Daily Fuel, Gold and Silver Prices and Latest Stories — On NDTV Profit.
