
transcript
show notes
OpenAI 規劃下一輪內部資安衝刺,邀 Cathedral Systems 的外部研究人員用 OpenAI 的模型測試自家防禦,重大等級漏洞獎金同時調高到四倍以上,特別突出的發現含加碼最高 $172,000。
⭐ 文章深度讀:OpenAI 手握模型,為什麼還請外部研究人員來打自家防線?
→ https://heymaibao.com/openai-outside-researchers-test-defenses-442dda/
⚡ 章節重點
先打自己的找洞競賽 00:00
OpenAI 的資安衝刺 00:55
漏洞獎金的新算法 01:27
受信任防守方的界線 02:29
個人版的先打自己 02:58
📝 懶人包
∙ OpenAI 規劃資安衝刺,找 Cathedral Systems 合作。
∙ OpenAI 宣布調高漏洞獎金,重大等級提高到四倍以上。
∙ OpenAI 說特別突出的發現,含加碼最高可拿 $172,000。
∙ 我的觀點:連做模型的公司找自家弱點,也沒有只靠模型和自己人。
📚 參考資料
As AI becomes more capable at finding vulnerabilities, defenders need to put it
→ https://x.com/OpenAI/status/2117444048227123277
OpenAI teams up with Cathedral to find and fix AI flaws faster, boosts …
→ https://www.indiatoday.in/technology/news/story/openai-teams-up-with-cathedral-to-find-and-fix-ai-flaws-faster-boosts-bug-bounties-3031143-2026-11-03
OpenAI raises top bug bounty payout to $172,000 with bonuses
→ https://www.helpnetsecurity.com/2026/11/03/openai-bigger-bug-bounty-rewards/
- https://heymaibao.com/openai-outside-researchers-test-defenses-442dda/heymaibao.com
- https://x.com/OpenAI/status/2117444048227123277x.com
- https://www.indiatoday.in/technology/news/story/openai-teams-up-with-cathedral-to-find-and-fix-ai-flaws-faster-boosts-bug-bounties-3031143-2026-11-03indiatoday.in
- https://www.helpnetsecurity.com/2026/11/03/openai-bigger-bug-bounty-rewards/helpnetsecurity.com