SECURITY TEAMS USED FRONTLINE AI TO SIMULATE ATTACKS ON 150 BITCOIN CORE PROJECTS, WITH AN AVERAGE OF ONE SERIOUS GAP PER HOUR PER PERSON

On 9 August, a volunteer safety team recently scanned some 150 Bitcoin core project-related code libraries using the Frontline AI model, and more than a dozen loopholes involving wallets, password libraries and infrastructure projects were identified. It is known that the team used Kimi K3 and OpenAI GPT Sol, Anthropic Claude Fable and Opus models and Z.ai GLM 5.2 to identify loopholes and generate supporting documents. Team members stated that there was currently an average of about one serious loophole per hour per person and that security reports had been submitted to several projects over the past 12 hours, but no specific projects affected had been made public. Recent security incidents such as Coldcard and Boltz also show that AI is being used by both security researchers and assailants to detect software loopholes more quickly。