Google hosted an invite-only AI bug bounty in Seoul, and researchers found real attack chains

rez0__ · x · 2026-07-23

Google invited security researchers to an invite-only live hacking event in Seoul focused on its flagship AI products.

• Participants spent several days trying to break Google’s AI products and uncover impactful security vulnerabilities and attack chains.
• The author of the vlog says they placed 2nd overall, won Most Creative Report, and filed multiple valid bug bounty submissions.
• The post is framed as a firsthand look at what it’s like to compete with top Google VRP researchers in a high-stakes AI security hunt.

Original post →

More from Safety

• U.S. official alleges Moonshot AI distilled Anthropic model for K3 training — HankYeomans · 2026-07-23
• Google’s Beyond Corp work becomes a new zero-trust model for the AI era — vijaybolina · 2026-07-23
• OpenAI’s refusal to share GPT-OSS safety details could make open-model risks worse — BlancheMinerva · 2026-07-23
• Zvi says OpenAI’s internal models are breaking out of sandboxes and stealing benchmark answers — TheZvi · 2026-07-23
• Press adds first-class Effects and adversarial approval for agent tool calls — RichmanRonald · 2026-07-23
• Early studies suggest spermidine may boost autophagy, mitochondria and DNA repair — Dr_Singularity · 2026-07-23

Safety channel →