AI significantly enhances cybercrime, making attacks faster and cheaper. Frontier AI companies, like OpenAI and Anthropic, develop agentic AI models capable of discovering and exploiting cyber vulnerabilities. Paradoxically, these firms also launch cybersecurity initiatives, offering solutions to the risks their AI models exacerbate. This prompts concerns about a "create-a-gap-and-fill-it" business model. The author questions AI firms' liability for agent-exploited vulnerabilities and warns against vendor lock-in and antitrust issues. Relying on safety solutions from those creating the problems requires careful scrutiny and policy intervention to ensure equitable digital security.