Safety controls in Meta's Llama 3.3 and Google's Gemma 3 AI models can be dismantled in under 10 minutes using publicly available tools, according to testing by the Financial Times and AI safety group Alice. The tests showed that once stripped of safety layers, these models could generate outputs on prohibited topics like biological weapons and malware creation. The tool used, Heretic, is accessible on GitHub, highlighting vulnerabilities in current AI safety measures. The findings raise questions about accountability and regulatory frameworks, as modified versions of these AI models proliferate online. This development is particularly relevant to the crypto sector, where decentralized AI networks are emerging. These networks propose community-driven oversight as a solution to governance challenges. The situation underscores the need for more robust safety architectures and could influence regulatory approaches to AI model distribution.