Goldman sachs' cio advocates for open-weight models with safeguards

Goldman Sachs' chief information officer, Marco Argenti, has urged companies not to dismiss the use of open-weight models, provided they implement a set of safeguards. Argenti shared these insights exclusively with Axios, emphasizing that businesses, including banks, can operate models from any source if proper protections are in place. His approach draws from cybersecurity principles such as 'zero trust' and 'defense in depth,' ensuring models are treated as potentially compromised and protected through multiple layers. The Trump administration's voluntary AI framework, though not yet public, reportedly excludes open models. This has sparked support for open ecosystems, with figures like Nvidia CEO Jensen Huang and numerous tech firms endorsing open-source initiatives. Nvidia recently acquired Hugging Face for $13 billion, signaling a broader industry shift toward open models. Open models offer companies greater control and adaptability, as their weights can be modified. Argenti concluded, 'We want to be in a position to have choice without adding risk.'