According to a16z, PANews states that companies face challenges with permissions and verifiability when deploying AI Agent. Encryption teams can leverage their experience with signatures and delegation for entry into this field.
According to PANews, a Safe multi-signature wallet is suspected of having had 6450000 USD stolen. Zodiac co-founder said it is very likely that the key used in the transaction has been leaked. Pantera stated that attackers have bought Monero tokens related to Hyperliquid.
PANews published an article exploring how the crypto industry can respond to attacks targeting employee identities, internal permissions, and supply chains in the AI era.
Replit founder predicts that enterprises will have humans set goals and AI Agent execute, while launching Agent Habitat to centrally manage permissions, approvals, and activity logs.
Anthropic Head of Platform Engineering Katelyn Lesse discusses the challenges of state, permissions, tracking, and emergency stops for long-running multi-Agent agents. The report also mentions a previously disclosed Anthropic incident involving unauthorized access in a test environment.
Beating reports that Cohere released an enterprise AI work platform, North 2, adding cross-session memory, shared knowledge, and Skills, fine-grained permissions and usage budget controls, while supporting cloud or private deployment.
OpenClaw Foundation released the Enterprise edition OCE, supporting enterprise self-deployment and centralized management AI Agent. OpenAI and Red Hat have conducted internal trials, but the project remains in the pre-production pilot stage.
OpenAI June apologized for unauthorized access to Australian government websites, pledging to establish a remediation working group, fund Australian cyber defense, and help address AI risks.