Google won’t tell anyone why its withholding Bard AI from the EU
While Google’s AI assistant Bard (opens in new tab) is currently available in 180 countries (opens in new tab) across the globe, the European Union and Canada still aren’t invited to the AI party. Almost two months after Google launched its friendly AI chatbot, Bard, the company is still withholding access to certain regions, but there’s no official statement on the matter.
The best guess is that Google may not see eye-to-eye with certain incoming regulations, not to mention that up against current GDPR rules, its processes may already be a little bit illegal.
The EU’s incoming AI Act (opens in new tab) is currently making its way through European Parliament in a bid to push current and would-be AI developers into making their products more transparent, and safer for the general public. Having spoken to some experts on the matter, Wired (opens in new tab) seems to be under the impression that Google is out there silently stomping its feet over the details of the act.
Even in its current state, Bard doesn’t quite fit the bill when it comes to the EU’s laws surrounding internet safety. As Access Now senior policy analyst, Daniel Leufer, says in the Wired piece, “There’s a lingering question whether these very large data sets, that have been collected more or less by indiscriminate scraping, have a sufficient legal basis under the GDPR.”
Aside from current law, the much more targeted, and rigorous AI Act set to pass in mid June would likely have a significant impact on how Google’s AI tool operates.
Once the bill goes through there will be even more restrictions placed on tools that could be “misused and provide novel and powerful tools for manipulative, exploitative and social control practices,” as is outlined in the official AI Act (opens in new tab) proposal. There are special mentions for specific human rights, such as the right to human dignity, respect for private and family life, protection of personal data, and the right to an effective remedy… all of which and more will be considered when dubbing an AI “high-risk.”
Looking at the AI tools of today, I’m having trouble thinking of any that don’t have the potential to encroach on at least one of those rights. It’s a scary thought, but it makes sense as to why Google might have some issues when it comes to Bard.
After all, as The Register (opens in new tab) notes, Italy, Spain, France, Germany, and Canada have all got their eye on ChatGPT (and presumably a bunch of other AI-based tools) over privacy concerns when it comes to user data. Canada’s AIDA proposal (opens in new tab), which will “come into force no sooner than 2025” explicitly calls for transparency in AI development, too.
Google’s AI principles (opens in new tab) state that it will not pursue the following:
- Technologies that cause or are likely to cause overall harm. Where there is a material risk of harm, we will proceed only where we believe that the benefits substantially outweigh the risks, and will incorporate appropriate safety constraints.
- Weapons or other technologies whose principal purpose or implementation is to cause or directly facilitate injury to people.
- Technologies that gather or use information for surveillance violating internationally accepted norms.
- Technologies whose purpose contravenes widely accepted principles of international law and human rights.
It’s a short list, and one with a few grey areas such as the use of “widely” and “internationally accepted norms”. Whether the backend could one day fully align with EU and Canadian law is unclear, but the language here could be a subtle way of leveraging a little wiggle room.
So, could it be that Google is trying to make a point by withholding Bard? Potentially.
Nicolas Moës, The Future Society’s European AI governance director, seems to think it possible. According to Moës, Google could well be making an attempt to “send a message to MEPs just before the AI Act is approved, trying to steer the votes and to make sure policymakers think twice before trying to govern foundation models”. Moës also notes that Meta has decided to withhold its AI chatbot, BlenderBot, in the EU too. So it’s not just Google playing it safe (or dirty).
It could also be that the big boys are keeping their toys to themselves because getting sued isn’t much fun. Either way, until Google comes out with an official statement, Europeans and Canadians alike will be left staring wistfully at Bard’s list of available countries.
