BNP Paribas analyst Karl Ackerman estimated that Nvidia's OpenShell architecture could cause a runtime slowdown of as much as 25%. Ackerman made the estimate in a note after GTC Berlin. The estimate does not establish performance across every task or deployment.
Nvidia recently introduced the OpenShell reference architecture at GTC Berlin. OpenShell builds on open-source software that controls agents' access and permitted actions. The system aims to constrain agents during operation, track actions and prevent activities that breach security rules.
AI agents can access files, use workplace tools and complete tasks with reduced human oversight. Agents may also connect to online services or execute code. Those permissions create risks beyond inaccurate chatbot answers if an agent misunderstands instructions or is manipulated.
Nvidia places security controls outside the agent instead of depending on the agent to regulate behavior. The architecture combines AI computing systems, BlueField-4 data processing units and Nvidia's Sentry security framework to monitor activity and apply restrictions. Nvidia said Cisco Systems, CrowdStrike Holdings, Microsoft and Alphabet's Google Cloud are working with the architecture.
Additional security checks can require more processing before an agent receives permission to act. Businesses must weigh the safeguards against possible delays, especially for urgent work or thousands of simultaneous tasks. A controlled agent with slower performance may nevertheless serve an enterprise better than a faster agent that accesses unauthorized information or performs unauthorized actions. The balance will vary with the task, security configuration and implementation.
OpenShell forms part of Nvidia's effort to offer products beyond AI chips. Security could account for an increasingly important share of enterprise AI spending as companies give assistants access to operational business systems. Large-scale adoption of Nvidia's approach remains a key question for investors. Another question is whether the controls can protect agents without substantially reducing performance. Independent benchmarks and customer deployments will be more important than the announcement alone. Security could give businesses another reason to select Nvidia's wider platform if Nvidia improves agent safety without materially slowing agents.
Nvidia Analyst Warns AI Safety Measures May Slow Agents by 25%
BenzingaShare
Nvidia's OpenShell may face support and resistance from traders concerned about a 25% runtime slowdown. BNP Paribas analyst Karl Ackerman cited FLASH AI data after the GTC Berlin launch. The architecture uses BlueField-4 DPU and Sentry security to control AI agent actions. Partners include Cisco, CrowdStrike, Microsoft, and Google Cloud. While security is key, the risk-to-reward ratio for urgent tasks remains unclear. OpenShell is part of Nvidia’s push beyond AI chips into enterprise security.
Source:Show original
Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information.
Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.