Joel Daroo, Agent Security at OpenAI, states that the rapid emergence of new AI capabilities in areas like cyber operations and message board manipulation caught security teams off guard. He explains that building a strong security posture requires more than just hardening systems; it demands that the people within an organisation change their habits and evolve alongside the technology. The speed at which these model abilities appeared created an extremely difficult problem for existing defences. Daroo now urges every organisation to assess how their staff, systems, and processes handle sudden jumps in AI capability. He asks whether teams know exactly what to do when something goes wrong or if they have the correct incident response plans ready. The core issue is whether an organisation has the right people available to act immediately when capabilities shift unexpectedly.
- Security culture must evolve faster than system hardening alone.
- Teams need clear protocols for sudden capability jumps.
- Incident response plans must account for unexpected AI behaviour.