Mitchell Amador, CEO of Immunefi, spent $10,000 in API credits for AI models to search for vulnerabilities in Bitcoin’s code. Research published by Anthropic in December 2025 demonstrated that AI agents can discover smart-contract vulnerabilities with potential exploit profits of around $2,500 per finding, making this type of investment profitable. Coinbase reduced its critical vulnerability rewards from $50,000 to $15,000 on July 29, 2026, a 70 percent decrease, due to the multiplication of AI-generated submissions overwhelming bug bounty programs. Smaller projects face an asymmetric threat: while a $10,000 AI-powered security audit is feasible for a well-capitalized firm, attackers face no such budget constraints when potential exploit gains far exceed the cost of the AI credits needed to prepare the attack.
Source: Read the original article
