Scientists urge tighter AI controls to prevent global cyber risks | Technology News | Al Jazeera

Over 300 scientists from 32 countries have signed a letter calling for stronger security measures in AI development to prevent models from going rogue and damaging critical infrastructure. The letter urges organizations developing advanced AI capable of cyberattacks to operate in tightly controlled environments, including air-gapped systems disconnected from the internet. Anthropic warned that its AI models could soon produce self-spreading malware affecting everyday internet-connected devices like smart TVs and appliances, with potential for autonomous, hard-to-contain infections.