Anthropic CEO Dario Amodei said that the speed at which the capabilities of artificial intelligence models are developed must now be managed more carefully. According to Amodei, as technology advances rapidly, security mechanisms cannot develop at the same pace, and this creates serious risks in controlling increasingly capable systems. Amodei specifically emphasized in his article that he did not advocate completely stopping artificial intelligence. Rather, he wanted progress to continue but be “adjusted” to give researchers time to develop security measures.
INDEPENDENT AUDITORS
The first leg of Amodei’s plan is independent security assessors. Accordingly, AI companies should allow outside experts to work within the company and directly examine the security practices of the models. These evaluators will not only review the reports prepared by companies; The company will be able to audit systems, model development processes and security incidents that may arise, with access close to the level that employees have. Anthropic has committed to implementing this application in-house.
COMMON STANDARD
In the second stage, companies developing artificial intelligence are expected to agree on common security standards. Amodei recommends that companies, especially those operating in democratic countries, set common rules regarding how quickly artificial intelligence models can be developed. The main purpose of this approach is to prevent competitors from gaining an advantage by speeding up if a single company slows down for security reasons.. According to Amodei, companies must compete with each other while also adhering to common security boundaries.
GLOBAL COOPERATION
The most difficult phase of the plan is international cooperation. Amodei argues that it will not be enough for companies in the USA and Europe to slow down, and that other countries affecting the development of artificial intelligence should also be included in the process.. In this context, it is recommended to try to agree on certain security rules with authoritarian governments such as China, as well as democratic countries. Amodei gives the example of international bans, especially in high-risk areas such as the development of biological weapons.
THE PURPOSE IS NOT TO STOP
Amodei also maintains his view that artificial intelligence can provide great benefits for humanity. Stating that we are talking about a technology that can provide significant advances in many areas, from the treatment of diseases to economic growth, Anthropic CEO argues that the main issue is to control the risks without losing this potential.
