The US government may weaken the cyber liability of artificial intelligence developers for a year. This approach was proposed by the AI Task Force: the relevant departments were advised to soften the application of sanctions on security incidents for at least 12 months.
The authors of the initiative expect that companies will become more likely to disclose information about failures and vulnerabilities if reporting a problem does not automatically result in a penalty. This should make the exchange of information more profitable than silencing the event. For industry, this is an attempt to obtain more risk data without immediately tightening controls.
The proposal marks a significant turn in American policy. In the past, authorities have threatened to apply existing legislation to the creators of unsafe AI products. Now the emphasis shifts to cooperation with the industry, although the threat of cyber incidents itself does not disappear. The main question is whether the temporary relaxation will lead to the disclosure of problems or will only reduce the responsibility of the entrepreneurs.