The Financial Post reports in its Saturday, June 13, edition that software tools are removing safety protections from artificial intelligence models by Meta, Google and other tech groups. A Financial Times dispatch to the Post reports that tests revealed that these modified AI systems responded to prompts about biological weapons, malware and child exploitation. Google's open-source model Gemma 3 provided instructions for dispersing chlorine gas, generated code for stealing credit card information and created stories about child sexual abuse. The FT used Heretic to strip the guardrails from Meta's Llama 3.3 model in under 10 minutes without specialized hardware. The modified model could address prompts the original system avoided, including lethal dosages of ricin. Concerns may increase among policy-makers and AI companies that enforcing safeguards will be more difficult as open-source systems become more powerful. The problem has intensified as frontier AI systems display increasingly sophisticated capabilities. The rise of modified models is hindering governments and AI companies in regulating systems during development, as downloadable tools can be copied and altered beyond their original creators' control.