Bill Gates, one of the co-founders of Microsoft, thinks that rapidly developing artificial intelligence technologies should be supported by stricter security mechanisms and regulations. Participating in NBC News’ Meet the Press program, Gates said that every artificial intelligence system should be required to have security measures and monitoring capabilities. Gates called not requiring such measures “completely irresponsible.” The statements came at a time when executives of leading artificial intelligence companies stated that the speed of development of the technology and the risks it brings should be considered more closely. According to Gates, it is necessary to go beyond voluntary measures left to companies’ own practices and apply certain security standards throughout the industry.
The approach of evaluating the security of artificial intelligence models through independent organizations is also becoming more popular in the technology sector. An example of this in the source text is Anthropic’s work with Accenture. Anthropic uses Accenture for third-party evaluation of the latest AI models. Such an approach allows developer companies to create an external evaluation layer in addition to their own security tests. However, Gates’ statements indicate that auditing should not be left solely to the choice of companies and that public authorities should also play a role in determining the rules to be applied.
Bill Gates draws attention to malicious uses in artificial intelligence
One of the options discussed regarding artificial intelligence security is a “kill switch”, that is, an emergency stop mechanism that can stop the operation of advanced systems when necessary. California Governor Gavin Newsom introduced this method as one of the elements of a more comprehensive security approach. Gates said during the interview that he was not categorically against such a mechanism. However, he stated that the emergency stop option alone would not solve the use of malicious artificial intelligence, which he considered a more imminent danger. According to Gates, the issue that needs more attention under current conditions is that malicious people can use advanced artificial intelligence tools for purposes such as bioterrorism or large-scale financial fraud.
In his Meet the Press interview, Gates stated that the most dangerous situation encountered today is “malicious people using artificial intelligence.” The former Microsoft executive also argued that the combination of malevolence and the latest AI tools could create a threat as powerful as any ever encountered. This assessment focuses AI security discussions on the possibility of currently accessible systems being misused by humans, rather than solely on long-term scenarios such as models getting out of control. In particular, facilitating access to information about biological threats and scaling financial crimes are among the risk areas that Gates focuses on. For this reason, he thinks that security mechanisms should include not only the technical behavior of the artificial intelligence model, but also methods that can monitor who uses the system and for what purposes.
Regarding legal regulations, Gates supports the involvement of law enforcement agencies and politicians in security discussions with artificial intelligence companies. According to him, regulatory bodies can play a role in determining what security and monitoring mechanisms are required from companies. Accepting that such a system would bring additional cost and operational burden to the artificial intelligence industry, Gates said that he did not think that this would lead to a “dramatic slowdown” in the work of companies. In other words, Gates advocates making certain security conditions standard for the industry, rather than completely braking artificial intelligence development activities. This approach requires striking a balance between rapid model development and oversight to reduce abuse risks.
For AI companies, independent testing, usage monitoring systems, and limiting access to risky talent are increasingly discussed. However, the scope of the security rules to be applied, which institutions will supervise them, and how they will be applied to artificial intelligence developers operating in different countries are still issues that need to be resolved. Gates’ statements are also based on the view that security standards, monitoring and public inspection should be considered together rather than a single technical measure. While an emergency stop mechanism can provide additional protection in certain scenarios, preventing malicious users from taking advantage of existing artificial intelligence tools requires different control methods. Therefore, the focus of the debate is how to implement rules that can reduce the risks of serious abuse without completely slowing down the development of artificial intelligence.