Topic: backdoor

A security researcher discovered that open-source machine learning models can contain hidden backdoors activated at a later date, raising concerns about potential future vulnerabilities in AI systems. This finding highlights the importance of thorough security audits for open-source projects to prevent unauthorized access or manipulation.