Ex-DOGE AI startup breaches Anthropic systems in 12 hours using AI agents
The breach exposed sensitive employee data, demonstrating how AI agent tools can be weaponized to compromise even top-tier AI security postures in under half a day.
The breach exposed sensitive employee data, demonstrating how AI agent tools can be weaponized to compromise even top-tier AI security postures in under half a day.
techstartups.com
The breach exposed sensitive employee data, demonstrating how AI agent tools can be weaponized to compromise even top-tier AI security postures in under half a day.
The incident highlights critical risks of failing to rotate exposed API keys, allowing attackers to compromise the popular LLM proxy tool's supply chain.
Payloads between 12.9 and 13.7 MB were large enough to bypass GitHub's code search, enabling widespread supply chain compromise of developer projects.
Redmond joins other maintainers scrambling to fix CVE-2026-111497, a compression library bug that enables remote code execution across affected runtimes.
The Claude maker's filing includes stark warnings that its models could blackmail humans, resist shutdown, or go rogue, as it prepares for one of the largest tech IPOs in history.
The move to cut ~60,000 roles comes as the tech giant pours massive resources into AI development, reshaping its operational structure.
Elon Musk says the funds will support AI infrastructure buildouts, while warning that smaller neocloud providers face collapse amid soaring AI chip demand.