A week before the new rules on General Purpose Artificial Intelligence (GPAI) come into effect, tools like ChatGpt and Gemini – clearly pictured the company stands when it comes to registering with the EU’s voluntary code of practice on GPAI.
Last week, Meta, the big tech giant in the US, said it would not sign the rules for suffocating innovation.
The code released last week by the European Commission is a voluntary set of rules that touch on issues of transparency, copyright, safety and security aimed at helping providers of GPAI models comply with AI laws.
It is expected that providers who sign up will be able to comply with AI laws and predict more legal certainty. Others will face more testing.
This is who is and who is out.
What to sign
US AI provider Anthropic, which developed AI assistant Claude as a competitor to Openai’s ChatGPT and Google’s Gemini, is the latest company that said it was going to sign the code.
“We believe the code is moving forward with principles of transparency, safety and accountability, a value that has long been defended by humanity in frontier AI development,” the company said in a statement.
“If implemented thoughtfully, EU AI laws and code will allow Europe to utilize the most important technologies of our time to power innovation and competitiveness,” the statement added.
Openai said last week earlier, claiming that Europe should “use this moment to enhance innovation and strengthen innovators to build for Europe’s future.”
The code drafting process that began last September after the committee selected a group of experts was heavily criticized by right-handed sholders, who feared violations of copyright laws, but the US tech giant claimed that the rules had curtailed innovation.
Microsoft President Brad Smith told Reuters last week that his company is likely to sign it as well. Smith said earlier this year that Microsoft wants to be a “voice of reason” as geopolitical tensions grow.
Things that don’t sign
The US tech giant Meta is the first and so far the only company that has said it will not sign the code. “Europe is on the wrong path for AI,” Joel Kaplan, chief executive, said in a statement last Friday.
After “carefully reviewing” the code, Meta will not sign because it “introduces a lot of legal uncertainty in model developers as well as measures that go far beyond the scope of AI laws.”
Gry Hasselbalch, an academic working on data and AI ethics and a contributor to the EU’s AI ethics guidelines, told Eunews that the code will not bring about a real change in the way companies implement general purpose AI in the EU.
“Like meta, companies that decide not to sign code must adhere to AI laws. So signing code is just a form. They still need to read and understand how AI systems are considered general purpose AI systems and how transparency, copyright and security in AI laws mean.
She added that the AI Act itself, the rules that regulate AI systems and tools in response to the risks they pose to society, have “been tokens in the geopolitical battle.”
“The law was developed in a carefully designed and implemented democratic process to create legal certainty for EU AI developers and employers. In fact, most AI systems are subject to existing laws without additional legal obligations under AI law,” she said.
Meta still needs to comply with the AI Act obligations that begin applications on August 2nd.
Other big tech companies, including Amazon, Google, did not yet want to comment on whether they would sign.
Providers who already have a GPAI model on the market must sign by August 1, while other providers can sign up later, the committee said. On that same day, EU executives will publish a list of signatories.
This code requires approval by the EU member states, represented by subgroups of the AI Committee, and the committee’s own AI office.
