Akamai Agent Sandbox Targets AI Agent Security
Akamai has launched Agent Sandbox in beta, a WebAssembly-based environment built with the open-source Spin framework to isolate code written and run by AI agents. Akamai Agent Sandbox limits network access and caps computing resources and memory, aiming to reduce data leaks and runaway code. The company also introduced Akamai Inference, a beta managed service for 12 open-weight AI models, with OpenAI-compatible APIs and serverless or dedicated GPU options. Both products are part of Akamai AI Cloud.
The launch comes as businesses give AI agents greater access to software, accounts and credentials. A 2026 Palo Alto Networks survey cited in the report found an average of 109 machine identities per human, including 79 AI agents, and organisations expect agent numbers to grow 85% over the next year. Risks include indirect prompt injection, stolen credentials and agents bypassing restrictions. The announcement adds to wider investment in AI security by companies including Palo Alto Networks, Cisco, CrowdStrike, Okta and Zscaler. It has no direct cryptocurrency or blockchain catalyst, though it reflects growing demand for AI infrastructure and security.
Neutral
The announcement names no cryptocurrency, blockchain network or token, so it provides no direct catalyst for crypto prices. In the short term, any market reaction is likely to be limited to AI and cloud-computing stocks rather than digital assets. Over the longer term, stronger AI security infrastructure could support wider use of AI agents, including in crypto businesses, but the article provides no evidence of immediate adoption or revenue effects for a specific token. Broader risk appetite and crypto-specific indicators are therefore more likely to drive trading and market stability.