OpenAI president warns companies about AI-powered cyberattacks:Shares 10 ways to stay safe and to defend against online threats

AI is making cyberattacks faster, smarter and potentially more dangerous. As attackers begin using advanced AI tools to find weaknesses in computer systems, OpenAI president and co-founder Greg Brockman has warned companies that they cannot afford to wait before strengthening their cybersecurity. His warning comes after OpenAI disclosed an incident involving Hugging Face, in which AI agents used in internal testing reportedly escaped their controlled environment and later compromised systems on the platform. Brockman believes the incident offers a glimpse of how cyber threats could evolve as AI becomes more capable. The capabilities of a typical threat actor will evolve in upcoming months.
AI could change the cybersecurity race According to Brockman, conversations with organisations in recent weeks have shown that businesses understand the need to improve their cybersecurity.
But he says they need to move much faster. AI could play a role on both sides of the cybersecurity battle.
Attackers can use it to identify vulnerabilities more quickly, while companies can use the same technology to find, prioritise and fix security weaknesses before they are exploited. Brockman says organisations should start using AI as a defensive tool now, rather than waiting until attackers gain a bigger advantage. OpenAI shares 10 steps companies should take Brockman has outlined several measures that organisations can begin working on immediately to strengthen their security. 1. Get support from company leadership Cybersecurity should be treated as a priority across the organisation. Companies need leadership support, resources, and funding to improve their defences quickly. 2. Give security teams an AI agent Businesses should consider using dedicated AI agents to assist cybersecurity teams with routine security tasks and help them respond faster to potential threats. 3. Equip AI with security expertise An AI agent is more useful when it has access to specialised cybersecurity knowledge, tools, and the right organisational data. 4. Test your systems now Companies should actively test their own infrastructure to find weaknesses before hackers discover them. Regular security assessments can reveal vulnerabilities that need immediate attention. 5. Clear the vulnerability backlog Known security flaws should not be allowed to pile up. Companies need to prioritise and fix existing vulnerabilities before attackers can take advantage of them. 6. Build security into software development Security should not be checked only after a product is developed. Companies should include security reviews throughout the software development process. 7. Use AI to fix vulnerabilities AI-powered security tools can do more than identify weaknesses. Companies can also use them to help resolve vulnerabilities and speed up the process of securing their systems. 8. Automate security alerts Security teams often have to deal with a large number of alerts. AI can help sort, investigate, and prioritise these warnings so that teams can focus on the most serious threats. 9. Prepare AI-powered forensics Companies should prepare automated investigation tools before a serious cyber incident happens. This can help security teams understand what went wrong and respond more quickly during an attack. 10. Keep experimenting Cybersecurity strategies need constant improvement. Brockman recommends security exercises, hack weeks and other experiments to test defensive systems and find better ways to protect organisations. Companies still have time to prepare Brockman’s warning is not just about the risks posed by AI. It is also a call for organisations to make use of the technology before cybercriminals gain a greater advantage. As AI systems become more capable, companies may need to automate more parts of their security operations and use AI to detect, investigate and fix threats faster.