According to MIT Technology Review, Anthropic released a report late last month arguing that the model GLM-5.3 poses a major cyberattack risk. The report has contributed to ongoing debates about whether the benefits of open AI models outweigh their potential security risks.
The publication frames this development within broader questions facing the AI industry about open model development. While open models offer transparency and accessibility benefits that have driven innovation in the field, concerns about malicious use cases continue to generate scrutiny from researchers and policymakers. The Anthropic report on GLM-5.3 specifically highlights cybersecurity vulnerabilities as a key risk factor.
MIT Technology Review is examining these trade-offs as part of their “MIT Technology Review Explains” series, which aims to help readers understand complex technological developments and their implications. The debate over open versus closed AI models remains a central tension in the field as capabilities continue to advance.