US legislators should take Sam Altman's warning seriously. He is the CEO of OpenAI and the creator of ChatGPT. Altman said that urgent regulation is necessary regarding AI due to the risks it poses to humanity.
"If this technology goes wrong, it can go quite wrong”, Altman told US senators at a Judiciary subcommittee hearing. They didn't seem entirely sure how urgent and complex a task they were given to solve.
The mere presence of the creators of an AI application that marked milestones in AI’s development at the end of last year means there is an effort to establish regulation in this area. The question was raised: How?
As global leaders in innovation, the US and the EU have already entered a field where they are looking for ways to regulate AI technologies, under pressure from their rapid development and growing fears that rules are lagging behind their progress.
The EU is ahead of the US in AI regulation
The EU is ahead of the US on this path. Adopting the AI Act, which has been ongoing for two years, is under way in the European Parliament. It has been frequently changed and improved precisely because of the tremendous speed with which technology moved ahead of regulation.
This will be the first law regarding the regulation of AI systems passed in the West, and is based on regulating different degrees of risk.
Divided into 4 groups, starting with unacceptable risk, whose application will be prohibited for use in the EU, to minimal or risk-free platforms.
Although it is the first of its kind, the common European legislation is already falling behind with decisions in some EU countries, where bans have already been introduced for some individual AI platforms, for example, in Italy, which has banned ChatGPT.
There are also different elements of regulation in the US. They are dispensed through different laws, for example, through laws on copyright, privacy, discrimination, and data protection, and at separate institutional levels - Congress and the administration.
Altman advocated the development of an agency, in the US or globally, before the Senate subcommittee. The agency would have the authority to issue licenses to the most powerful AI systems and to revoke them if it determines the abuse or risk of an AI platform concerning security standards.
Regulation lags behind AI development
It seems that the road to such a solution is still long. There is a constant risk of the evolution of AI technologies moving faster than regulations.
At the same time, the regulation has meaning only at the level where its power of control reaches, so it is an open question whether setting up a central regulator would make sense at this moment or in a year or 2 when AI systems will outgrow the current level of development.
There is a greater need for the US and the EU to harmonise their strategies as an area that will dictate the future global model. At this point, those paths have more differences than similarities.
Despite all the differences, the EU and the US have a common starting point. Regulating AI technologies should be based on a risk assessment which is a good basis for harmonising legislation.
The main effort for performing groundwork regarding AI’s global governance is going through the joint US-EU Trade and Technology Council.
In Japan, the G7 ministers for digital technologies also paid a lot of attention to the regulation of AI, that is, "responsible AI and global AI governance", whereby they accepted the risk-based approach represented by the EU.
Beat China in creating a model for regulation
A perfect agreement between the US and the EU, as the 2 largest technological and innovative global areas, is perhaps difficult to achieve but a joint approach to AI is realistic, at least in basic principles.
“I don’t think there will be perfect alignment. There never really is in terms of regulations at the international level. … But what there can be alignment on, for example, is to make sure we’re discussing the same thing… alignment on the key risks, on some of the key definitions, the key parameters, the key principles”, said William Long, head of the EU Data Protection group.
A joint approach is necessary because the race with China over the regulation of AI systems is ongoing. China has been rapidly developing its own models.
The Chinese models of AI regulation will aim to become globally accepted. This carries a risk of spreading a concept in the centre of which is state control and supervision over the development of AI.
“All the (Chinese) regulations so far, pretty clearly have their roots in fears about losing control over the flow of information. In terms of the Chinese government’s relationship to AI, it’s very broadly supportive but wanting to cover its bases on control of information”, said Matt Sheehan of the Carnegie Endowment for International Peace.
China's view of AI development is fundamentally authoritarian and represents a risk for data privacy, its misuse, and particularly for setting limits on AI development and innovation.
Due to this limitation, Western models of AI regulation have no reason to strive for global acceptance because, for their democratic solutions, they will always have an insurmountable obstacle in China's authoritarian vision of development and control over AI.
The Western model, or at least the common basic principles on which AI regulation would be based, would be sufficient as a template that could be accepted by many others outside the US and the EU, but without the imperative of becoming a global standard.
