OpenCode breach hits 1.4% of users, AI filters block defense — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenCode disclosed Saturday that attackers breached an older system within its Console platform, exposing API keys for 1.4% of users and prompting the team to email every affected account.
- Users posted overnight spending alerts firing across OpenAI, Anthropic, and Google accounts, with some describing unauthorized charges topping $1,000.
- Anthropic and Google blocked exposed API keys identified during the incident, while OpenAI worked with OpenCode to restore access and refund credits consumed in the attack, per co-founder Dax Raad.
- OpenCode ran AI defensive agents around the clock for 48 hours as attackers kept probing, but frontier models' safety filters began refusing to examine malicious payloads once code triggered guardrails.
- Hugging Face hit the same wall in July when commercial models refused to analyze intrusion logs from an autonomous agent attack — calling it the "asymmetry problem" where guardrails "cannot distinguish an incident responder from an attacker."
- David Sacks, co-chair of the President's Council of Advisors on Science and Technology, posted that "the guardrails backfired" because frontier models refused to analyze the exploits needed for defense.
- OpenCode received access to OpenAI's Daybreak Blue program for vetted defensive work after standard models refused; the breach occurred the same night the company made GLM-5.3-Flash free, a model researchers flagged for cybersecurity capabilities approaching restricted-access tiers.
Why it matters: The breach crystallizes a structural tension in AI safety: the same guardrails blocking malicious use also prevented OpenCode from analyzing the attack in real time, forcing it to seek Daybreak Blue access mid-incident. With unauthorized charges exceeding $1,000 for some users, incident responders now have a concrete case for vetted, less-restricted models — as white House science advisor David Sacks publicly argued.
Ask SkimNews
