ChatGPT has been around for ~6 months, and ChatGPT Plus (powered by GPT-4) for ~3 months now. After the release of GPT-4:
Microsoft announced that they have already started using it to power Bing, which in turn is available both through Microsoft's internet browser (Edge) and via Skype, too.
OpenAI announced the ChatGPT plugins, which allow the interconnection of GPT-4 and related products with other third-party applications.
Beyond the OpenAI plugins, a separate and independent ecosystem for interconnecting LLMs (including, but not limited to, GPT) with other applications has quickly emerged around LangChain
My impression is that, beyond Bing, the integration of GPT-related products with third-party applications through the OpenAI plugins and LangChain has been steadily growing. Also, besides concerns and worries, the tool seems to be immensely popular with practitioners (just check Ethan Mollick's substack and twitter feed).
On April 1, Italy banned ChatGPT in the country over privacy concerns, giving OpenAI 20 days to say how they would address these concerns. Following the Italian ban, there were rumors for other European countries following suit (namely France, UK, and Ireland). Something like that has not happened so far - in fact, Irish regulators argued against "rushing into prohibitions that really aren't going to stand up" . After the ban, Google searches for "VPN" skyrocketed in Italy; according to TechRound:
Google Trends data shows that searches for VPNs in Italy have skyrocketed since the announcement of the ChatGPT ban, indicating a strong desire among the population to continue using the AI chatbot despite the government’s concerns.
The surge in VPN usage raises questions about the ban’s effectiveness and whether it will ultimately protect user privacy or drive users to seek alternative methods to access ChatGPT. Some experts argue that the ban may have inadvertently increased the risk of privacy breaches as users turn to potentially unsecured VPN services to bypass the restrictions.
On April 18, the European Data Protection Board set up a ChatGPT task force. I have not located any further news or updates from them so far.
On April 28, after corrective actions taken by OpenAI, Italy announced that they lift the ban; according to The Verge:
Among the changes, OpenAI linked The Verge to a new form that EU users can submit to remove personal data under Europe’s General Data Protection Regulation (GDPR). It also says that a new tool will verify users’ ages upon signup in Italy, and it published a help center article that outlines how OpenAI and ChatGPT collect personal information, including information about contacting its GDPR-mandated data protection officer.
It would seem from the above that a good part of the corrections made for Italy are applicable to (and address concerns of) other EU countries as well.
Last week, OpenAI CEO Sam Altman went on a European tour, where he met with the leaders of Germany, France, Spain, Poland, UK, and the president of the European Commission. I take that a main concern of British PM Rishi Sunak, according to his own words, was "to discuss how the UK can provide international leadership on AI" (does not imply exactly a ban mood). Altman (among others) had also a teleconference with the (all too powerful) European commissioner for competition, Margrethe Vestager. (FWIW, notice that, except for the German chancellor, all the relevant tweets above are from the leaders' Twitter accounts - Altman just retweeted them...)
Altman is currently in Israel (he met/talked with both the president and PM), and this week is scheduled to travel to Jordan, Qatar, UAE, India, and South Korea (Reuters, Twitter):
"Israel - like Britain, and to a great extent like Canada, too - is at the U.S. end of the spectrum," Ziv Katzir, director of national AI planning at the Israel Innovation Authority, told Reuters.
How many countries have banned ChatGPT?
We have a not-updated table here (from February): It is neither updated nor fully accurate - Vietnam is missing, while Paraguay and Sao Tome & Principe are now in the list of supported countries. Adding Italy, we have that the number of countries where ChatGPT has not been available for whatever reason (ban or otherwise), even temporarily, is currently 34.
ASSESSMENT
At least in non-authoritative states, bans normally do not come suddenly and out of the blue; they need existing regulations to step upon (GDPR in the case of Italy), and usually they are preceded by a consultation phase. Italy's action seems to be an outlier here, and it's telling that all other EU countries, despite being under the very same regulation (GDPR), were very reluctant to follow suit. At least for Western countries, it would seem that the present situation is moving towards a "talk first" approach already suggested by the Irish regulators, which, on top of the OpenAI actions to address the Italian ban, seem to have made more bans from EU and similar countries less likely in the near future.
The actual replication of ChatGPT into Bing (with Bing in turn already integrated into the Microsoft Edge browser and Skype), which was still in pilot phase when Italy banned, would make a ban much more difficult in practice today. Add the bypass possibilities with VPNs and the gradual integration of ChatGPT with third-party applications that are already being widely used by practitioners, and the idea of a ban by a democratic non-authoritative state starts to sound like a bad joke...
Some fellow forecasters below have argued (implicitly or explicitly) that concerns over misinformation and/or unemployment could lead to a ban somewhere and in the time frame of the question. At least for democratic non-authoritative states, I think the probability of something like that is close to zero; such concerns are presently way too vague to justify a ban in non-authoritative states, and feasible candidate reasons would need to be solidly grounded on existing regulations (new regulations move at glacial speeds, and I don't expect any to be ready in the time frame of the question). Similarly, I consider banning ChatGPT due to fears of a forthcoming superintelligence to have a chance of exactly zero. For such countries, I cannot see any other realistic reason than data privacy & protection (copyright concerns currently seem too vague to me, too, and there are already some signs that the mood is going towards allowing copyrighted works to be used as training material for AI systems).
Notice that we should not read too much in Altman's recent meetings with European leaders - European regulators are largely independent of their respective governments (at least on paper); but they seem to signify a certain mood not toward bans. --- Authoritative states are a different story altogether (and much less predictable), but it would seem that a good portion of them (including all the "usual suspects", and then some) are already off the list of supported countries.
The unpredictability of authoritative states stems from the fact that we consider they can do more or less whatever they like, without accountability to their people, which can produce actions that may even seem irrational to an outside observer. Still, people do not do things just because they can, but because they have a reason to do so. And presently I confess that I have difficulty imagining why exactly any official in such a country still in the list would bother to ban something like ChatGPT, which does not seem to constitute any direct threat (of course, this may say more about my own lack of imagination, rather than about anything else...)
CONCLUSION & FORECAST
Is there a base rate? I will admit it is highly debatable, but I would argue that the percentage of countries that have ever been off the list for whatever reason (ban or other) is not an unreasonable base rate here; currently, with the number being 34 over 195 countries in the world, this percentage comes to 17.4%, pretty close both to my initial rough forecast of 15% and the current crowd consensus of 21.5%.
I will shave off this number to 14%, since this base rate is somewhat artificially inflated - it includes countries where OpenAI had either no possibility (due to US regulations) or no interest to provide access (think of Afghanistan, Iran, North Korea etc). I also expect that the crowd consensus will fall even more from its current 21.5%.
--- Feedback and comments (especially against my arguments & rationale) are always most welcome and actively encouraged.
Files
Why might you be wrong?
~165 countries in the list are an awful lot, and a single ban will trigger a Yes resolution. It might very well be the case that such an event is truly radically uncertain, in which case probabilistic assessments may be inadequate in the first place.
Why do you think you're right?
FACTS
ChatGPT has been around for ~6 months, and ChatGPT Plus (powered by GPT-4) for ~3 months now. After the release of GPT-4:
My impression is that, beyond Bing, the integration of GPT-related products with third-party applications through the OpenAI plugins and LangChain has been steadily growing. Also, besides concerns and worries, the tool seems to be immensely popular with practitioners (just check Ethan Mollick's substack and twitter feed).
On April 1, Italy banned ChatGPT in the country over privacy concerns, giving OpenAI 20 days to say how they would address these concerns. Following the Italian ban, there were rumors for other European countries following suit (namely France, UK, and Ireland). Something like that has not happened so far - in fact, Irish regulators argued against "rushing into prohibitions that really aren't going to stand up" . After the ban, Google searches for "VPN" skyrocketed in Italy; according to TechRound:
On April 18, the European Data Protection Board set up a ChatGPT task force. I have not located any further news or updates from them so far.
On April 28, after corrective actions taken by OpenAI, Italy announced that they lift the ban; according to The Verge:
It would seem from the above that a good part of the corrections made for Italy are applicable to (and address concerns of) other EU countries as well.
Last week, OpenAI CEO Sam Altman went on a European tour, where he met with the leaders of Germany, France, Spain, Poland, UK, and the president of the European Commission. I take that a main concern of British PM Rishi Sunak, according to his own words, was "to discuss how the UK can provide international leadership on AI" (does not imply exactly a ban mood). Altman (among others) had also a teleconference with the (all too powerful) European commissioner for competition, Margrethe Vestager. (FWIW, notice that, except for the German chancellor, all the relevant tweets above are from the leaders' Twitter accounts - Altman just retweeted them...)
Altman is currently in Israel (he met/talked with both the president and PM), and this week is scheduled to travel to Jordan, Qatar, UAE, India, and South Korea (Reuters, Twitter):
How many countries have banned ChatGPT?
We have a not-updated table here (from February):
It is neither updated nor fully accurate - Vietnam is missing, while Paraguay and Sao Tome & Principe are now in the list of supported countries. Adding Italy, we have that the number of countries where ChatGPT has not been available for whatever reason (ban or otherwise), even temporarily, is currently 34.
ASSESSMENT
At least in non-authoritative states, bans normally do not come suddenly and out of the blue; they need existing regulations to step upon (GDPR in the case of Italy), and usually they are preceded by a consultation phase. Italy's action seems to be an outlier here, and it's telling that all other EU countries, despite being under the very same regulation (GDPR), were very reluctant to follow suit. At least for Western countries, it would seem that the present situation is moving towards a "talk first" approach already suggested by the Irish regulators, which, on top of the OpenAI actions to address the Italian ban, seem to have made more bans from EU and similar countries less likely in the near future.
The actual replication of ChatGPT into Bing (with Bing in turn already integrated into the Microsoft Edge browser and Skype), which was still in pilot phase when Italy banned, would make a ban much more difficult in practice today. Add the bypass possibilities with VPNs and the gradual integration of ChatGPT with third-party applications that are already being widely used by practitioners, and the idea of a ban by a democratic non-authoritative state starts to sound like a bad joke...
Some fellow forecasters below have argued (implicitly or explicitly) that concerns over misinformation and/or unemployment could lead to a ban somewhere and in the time frame of the question. At least for democratic non-authoritative states, I think the probability of something like that is close to zero; such concerns are presently way too vague to justify a ban in non-authoritative states, and feasible candidate reasons would need to be solidly grounded on existing regulations (new regulations move at glacial speeds, and I don't expect any to be ready in the time frame of the question). Similarly, I consider banning ChatGPT due to fears of a forthcoming superintelligence to have a chance of exactly zero. For such countries, I cannot see any other realistic reason than data privacy & protection (copyright concerns currently seem too vague to me, too, and there are already some signs that the mood is going towards allowing copyrighted works to be used as training material for AI systems).
Notice that we should not read too much in Altman's recent meetings with European leaders - European regulators are largely independent of their respective governments (at least on paper); but they seem to signify a certain mood not toward bans.
---
Authoritative states are a different story altogether (and much less predictable), but it would seem that a good portion of them (including all the "usual suspects", and then some) are already off the list of supported countries.
The unpredictability of authoritative states stems from the fact that we consider they can do more or less whatever they like, without accountability to their people, which can produce actions that may even seem irrational to an outside observer. Still, people do not do things just because they can, but because they have a reason to do so. And presently I confess that I have difficulty imagining why exactly any official in such a country still in the list would bother to ban something like ChatGPT, which does not seem to constitute any direct threat (of course, this may say more about my own lack of imagination, rather than about anything else...)
CONCLUSION & FORECAST
Is there a base rate?
I will admit it is highly debatable, but I would argue that the percentage of countries that have ever been off the list for whatever reason (ban or other) is not an unreasonable base rate here; currently, with the number being 34 over 195 countries in the world, this percentage comes to 17.4%, pretty close both to my initial rough forecast of 15% and the current crowd consensus of 21.5%.
I will shave off this number to 14%, since this base rate is somewhat artificially inflated - it includes countries where OpenAI had either no possibility (due to US regulations) or no interest to provide access (think of Afghanistan, Iran, North Korea etc). I also expect that the crowd consensus will fall even more from its current 21.5%.
---
Feedback and comments (especially against my arguments & rationale) are always most welcome and actively encouraged.
Why might you be wrong?
~165 countries in the list are an awful lot, and a single ban will trigger a Yes resolution. It might very well be the case that such an event is truly radically uncertain, in which case probabilistic assessments may be inadequate in the first place.