Home | Newsletters & Shorts | Digital Watch newsletter – Issue 89 – May 2024

Digital Watch newsletter – Issue 89 – May 2024

Download your copy

 Person, Head, Face, Boxing, Sport

Snapshot: The developments that made waves

AI governance

South Korea announced its plans to host the 2nd global AI safety summit. Meanwhile, the UN Development Programme (UNDP) facilitated discussions between the G7 Italian presidency and the African Union, focusing on leveraging AI for sustainable development. Japan’s Prime Minister Kishida Fumio announced that over 49 countries and regions have joined the Hiroshima AI Process Friends Group. 

In a collaborative effort with Germany and the EU, Kenya launched a project to develop its National AI Strategy. 

In the USA, the National Security Agency’s Artificial Intelligence Security Center (AISC) issued new guidance on securing AI systems. In a related development, the National Institute of Standards and Technology (NIST) issued four draft reports aimed at enhancing AI system standards. Meanwhile, in Italy, the Cassa Depositi e Prestiti (CDP) announced a substantial investment of EUR1 billion in AI and cybersecurity.

The AI landscape saw a burst of innovation as industry giants OpenAI, Google, Meta, and Mistral each unveiled new large language models (LLMs). Anthropic also made headlines by claiming its AI models are now as persuasive as humans.


In a bid to reduce dependency on major global chipmakers, Google introduced a new AI chip. South Korea plans to invest USD6.94 billion in AI by 2027, along with a separate USD1.1 billion fund to foster AI semiconductor firms. Samsung also announced a USD44 billion investment in the USA to expand its chip manufacturing. China set a 2027 deadline for its telecom sector to replace American chips with local alternatives. The USA announced a global partnership aimed at enhancing cooperation in AI and semiconductors, which could be seen as a counterbalance to China’s growing influence in the sector. Read more about China’s and the USA’s battle for chip supremacy below.


The Belgian EU Council presidency unveiled the framework for an online child protection law.

Across the Atlantic, the US Cybersecurity and Infrastructure Security Agency (CISA) issued Emergency Directive 24-02 as a countermeasure against a Russian cyberattack that targeted Microsoft email accounts. In response to evolving cyber threats, the US Department of State introduced a new Cyberspace & Digital Policy Strategy to bolster national security and international cooperation in digital spaces.
The private sector also experienced disruptions, with Dutch chipmaker Nexperia reporting a cyberattack and RansomHub exposing personal and medical records taken from Change Healthcare during a cyberattack earlier this year. The UNDP confirmed that it had also suffered a ransomware attack.


The US Federal Communications Commission (FCC) has reinstated net neutrality rules and asserted new regulatory control over broadband internet, reversing the policies enacted under the former administration.
The Council of the European Union has given final approval to the Gigabit Infrastructure Act, a new law to accelerate the deployment of high-speed networks in the EU.


China plans to establish a national computing network with its National Data Administration to address regional digital gaps and enhance capabilities.

The US Congress introduced the Generative AI Copyright Disclosure Act, which requires AI companies to disclose copyrighted materials used in their training datasets to the Register of Copyrights before releasing new AI systems.

Internet economy

EU antitrust regulators are scrutinising a proposal by Apple to determine whether it allows Spotify and other music streaming services to inform users of alternative payment methods outside of Apple’s App Store. 

The UK treasury announced plans to introduce a comprehensive regulatory framework for cryptocurrencies and stablecoins. 
The Canadian government announced its plan to push ahead with a digital services tax starting in 2024. The tax specifically targets big tech companies operating within the country.

Digital rights

Federal legislators in the USA introduced a bipartisan bill titled the American Privacy Rights Act. The proposed legislation encompasses limitations on collecting personal data, increased consumer control over their personal information, and the implementation of protective measures against discriminatory algorithms.

The Ethiopian Parliament enacted comprehensive digital data protection legislation, specifying the rights of data subjects, establishing statutory obligations for entities that process personal data, and setting forth limitations on collecting and using such data.

Finally, NOYB, an Austrian privacy advocacy group, filed a complaint against OpenAI, alleging that it has failed to adequately address the dissemination of incorrect information about individuals by its ChatGPT.


The global landscape for internet freedom has experienced a setback, with 1st quarter of 2024 recording 22 intentional internet shutdowns in 12 countries, a recurrence of the peak observed in 2021. 

The South African government initiated a comprehensive plan to connect approximately 5.5 million households in rural communities and townships to the internet over the forthcoming 3–4 years.

In Brazil, a tripartite initiative involving Anatel, UNESCO, and the Brazilian government has been launched to close the digital divide by expanding internet access throughout the country, specifically targeting underserved regions.


Microsoft disclosed that Chinese groups were leveraging AI to create content aimed at distorting the US elections. 

The Supreme Court of Brazil initiated an investigation into X’s (formerly Twitter) actions regarding social media accounts. X agreed to comply with Brazilian court orders, a move possibly aimed at mitigating further legal complications. Pakistan blocked access to X, citing national security as the primary reason.

Meta announced a new policy to label AI-generated content rather than removing it. 

Elsewhere, an Italian senator proposed the establishment of a national agency dedicated to combating disinformation.


In April, the highlight of multilateral Geneva when it comes to digital governance was the 27th session of the Commission on Science and Technology for Development (CSTD). The session revolved around two key themes: data for development and global cooperation in science, technology, and innovation for development. It also addressed the progress in implementing the outcomes of the World Summit on the Information Society (WSIS), which gains importance ahead of the WSIS+20 review in 2025. The Economic and Social Council (ECOSOC) entrusted the CSTD  to review WSIS outcomes and prepare a WSIS+20 report. The CSTD presented a timeline of its preparation during the session. An agenda item dedicated to the Global Digital Compact (GDC), offering member states an opportunity to share their views of the GDC Zero Draft, was included. 

Speaking of the GDC, we can safely say it was the talk of the town in Geneva in April, as member states, UN agencies, and other stakeholders discussed the Zero Draft in various settings. Read more about the GDC process and WSIS+20 process on our Digital Watch Observatory.

China and USA box it out over tech

For a considerable span of time, China and the USA have found themselves at odds in the realm of technology. The focal points of contention in April? TikTok and chips.

The crux of America’s problem with TikTok is data safety: is US user data being regularly sent to China by TikTok? 

Just to remind you, TikTok’s parent company, ByteDance, is based in China, whose 2017 National Intelligence Law requires companies to assist with state intelligence work. This is where fears about the transfer of user data to China come from. 

TikTok has repeatedly claimed that China does not have access to US user data. Americans have never been very convinced of this, and their concerns skyrocketed again in April. 

The preferred solution for this dilemma, at least for the USA, is the divestment of TikTok, which was conveyed to China on the highest level – from US President Biden to China’s President Xi, in a phone call. 

In April, Biden signed into law a foreign aid package that includes a provision to ban TikTok if ByteDance fails to divest the app within a year. More precisely, ByteDance has an initial nine-month period to finalise a deal (until 24 January 2025), which could be extended for three more months if progress is evident. The first deadline is 19 January 2025, with the final deadline being 19 April 2025. 

But ByteDance is not contemplating a sale: TikTok’s algorithms align closely with the company’s domestic apps, making it challenging to divest without relinquishing crucial intellectual property. 

ByteDance and TikTok have instead decided to fight it out in a legal battle and are suing the US government over the law

ByteDance and TikTok argue that the US government has not presented sufficient evidence of the alleged Chinese misuse of the data associated with the app. The companies seek a court judgement to declare the legislation unconstitutional because it infringes upon the US First Amendment, which protects the freedom of speech. They have also requested a court order to prevent the attorney general from enforcing the law.

This situation is unusual, so it’s difficult to say how the court will rule. It does, however, have wider implications on data governance: those who control data will control the world.

Abstract digital waves with flowing particles.

The two countries are also trading blows over chips. It’s hard to say who threw the first punch – perhaps it was the USA when its Department of Commerce placed Fujian Jinhua Integrated Circuit Co Ltd on a list of entities that cannot buy components, software, or technological goods from US firms in 2018.

What can be identified as one of the most brutal punches was the US October 2022 export controls on advanced computing and semiconductor manufacturing items. 

These measures prevent US firms from selling advanced chips to China or providing Chinese firms with tools for making their own advanced chips. They also prevent firms from other countries from doing the same if those chips or tools were made with US-developed technologies. 

Why did this punch land so hard? There are almost no semiconductors without some kind of US-trademarked bits in their design or production processes.

This led to China initiating a trade dispute procedure at the World Trade Organization (WTO) against US chip export control measures, arguing that these measures ‘threatened the stability of the global industry supply chains’. 

In October 2023, the White House imposed a new export ban on advanced chips.

At the end of March 2024, these export controls tightened, a move that China blasted, citing, again, increased trade barriers and uncertainty in the global chip sector. 

However, Chinese entities managed to bypass these export bans and get their hands on some Nvidia chips through resellers, thus raising doubts about the effectiveness of export controls. Nvidia and Intel are also planning to produce China-specific chips, which will be developed in compliance with US export controls and sanctions. 

China also enforces its Document 79, a secretive policy nicknamed ‘Delete A’ or ‘Delete America’, which mandates vital sectors like finance and energy to phase out all foreign software. 

Both countries are ramping up investment in local production of chips. China is mobilising a USD27 billion chip fund called the National Integrated Circuit Industry Investment Fund. And Biden inked a deal with Micron Technology at the end of April, securing up to USD6.14 billion for two new chip factories in New York and Idaho. A very recent report suggests that the USA will have even more success: By 2032, the USA could produce 28% of advanced chips, while China is expected to produce only 2%.

And what of allies? China is calling for continued advanced chipmaking equipment trade with the Netherlands, while the USA is pushing its partners, namely the Netherlands, Germany, South Korea, and Japan, to impose stricter chip export controls. 

Oddly enough, it is Malaysia that is benefiting the most from these tensions, as both China and the USA are investing in chipmaking there

It also bears mentioning that the heavyweight in chipmaking is centred in Taiwan, which has a 61% market share in chipmaking and produces around 92% of the world’s most advanced chips. 

Is this a fight with no winners? Considering when the match began, how many more rounds are we in for? Could tensions ever simmer down? 

Some of that hinges on the outcome of this year’s presidential election in the USA – if former President Trump retakes the White House, the dynamics, at least for TikTok, could change, as he already indicated he is against the ban. 

In the meantime, the chips race is even more likely to accelerate due to the April earthquake in Taiwan, a stark reminder of how dependent our devices are on Taiwan’s chips.

 Scoreboard, Bar Chart, Chart

The intellectual property saga: Strategies for balancing AI advancements and IP protection | Part 3

We continue our in-depth exploration of AI’s influence on intellectual property. The first part delved into the intricate process of copyrighting AI-generated content, revealing the challenges posed by traditional ownership laws. 

The second essay navigated the complex terrain of AI’s influence on trade secrets and trademarks within the EU and US legal frameworks. 

This final part will further illuminate the multifaceted strategies implemented to safeguard IP in the Age of AI, underscoring the depth and significance of this evolving issue.

Shiny blue drawing of a complex circuit in the shape of a human brain

The legal landscape surrounding AI and intellectual property is intricate and rapidly evolving. It demands careful consideration of human and AI contributions to creative and innovative endeavours. 

While AI is a significant contributor to innovation, it faces hurdles in getting its input recognised. For instance, the US Court of Appeals recently rejected AI as an inventor, aligning with human-centric patent laws. Copyright registrations for AI-generated works are also contentious due to human authorship requirements. 

However, it’s important to note that ongoing efforts, as seen in cases like Kashtanova’s comic book, are being made to protect human-authored elements. Patenting AI systems under Alice Corp. v. CLS Bank International requires inventive transformations for eligibility, navigating a two-step test assessing subject matter and imaginative concept. This dynamic landscape underscores the urgency and importance of staying updated.

Distinguishing between inspiration and infringement becomes crucial, asking for the establishment of governance mechanisms to address these concerns and maintain trust within creative industries. The legal battle between the Writers Guild of America and the Alliance of Motion Picture and Television Producers is a stark reminder of the importance of regulating AI usage. While compromises have been reached to balance innovation and copyright protection, the stipulation that AI-produced content cannot be credited as literary material, but collaborative work with human writers is allowed, is a significant development. 

But is there a way to detect AI copyright infringements? Originality.AI reports that nearly 20% of the top 1,000 websites block crawler bots from gathering data for AI use. LLMs such as ChatGPT and Google’s Bard need extensive data for training. Tech providers offer AI-powered solutions for IP protection, using machine learning to detect infringement. These tools track copyrighted material across websites and social media, helping rights holders protect their IP rights.

As AI technology progresses, it raises questions about the ownership of AI-generated content. Who owns the rights to creations like AI-composed music or inventions? Current IP laws lack clarity on this, underscoring the necessity for reforms to accommodate AI’s role. Copyright and trademark laws also need adjustment for AI-generated works. 

Many innovators opt to use trade secret processes over patents to protect their AI advancements. Future regulations must address these issues to foster innovation. Implementing a distinct protection system for AI-generated creations and discussing algorithm protection are potential solutions. The effective regulation of AI-related IP rights is crucial for legal clarity and innovation.

A more in-depth version of this text will be available on the Digital Watch Observatory shortly.  Meanwhile, explore the previous parts of our AI and intellectual property series: The intellectual property saga: The age of AI-generated content | Part 1 and The intellectual property saga: AI’s impact on trade secrets and trademarks | Part 2.

AI and copyright

Patent authorities globally are grappling with the challenge of redefining their approach to handling inventions generated not by human ingenuity but by AI. It has sparked considerable debate within the intellectual property community.

double exposure creative human brain microcircuit with hand writing notepad background with laptop future technology ai concept

In continuing the three-part series on AI’s influence on intellectual property, the first part discussed the complexities of copyrighting AI-generated content, noting the challenges of traditional laws in ownership. The second section will explore AI’s impact on trade secrets and trademarks in the EU and US legal frameworks

Digital dominance in 2024 elections worldwide

As a historic number of voters head to the polls, determining the future course of over 60 nations and the EU in the years ahead, all eyes are on digital, especially AI.

Digital technologies, including AI, have become integral to every stage of the electoral process, from the inception of campaigns to polling stations, a phenomenon observed for several years. What distinguishes the current landscape is their unprecedented scale and impact. 

World map signals the 54 courntries with democratic elections during 2024, notably including many EU countries and the USA, as well as xxx [check third island area, I cannot tell which countries are highlighted
A map of elections in 2024 Image credit: Gateway house

Generative AI, a type of AI enabling users to quickly generate new content, including audio, visuals, and text, made a significant breakthrough in 2023, reaching millions of users. With its ability to quickly produce vast amounts of content, generative AI contributes to the scale of misinformation by generating false and deceptive narratives at an unprecedented pace. 

The multitude of elections worldwide this year, pivotal in shaping the future of certain states, have directed intense focus on synthetically generated content, given its potential to sway election outcomes.

Political campaigns experience the emergence of easily produced deepfakes, stirring worries about information credibility and setting off alarms among politicians who called on Big Tech for more robust safeguards.

Tech companies are, by and large, partnering with individual governments’ efforts to tackle the spread of election-related misinformation. 

But, some remain dissatisfied with the ongoing efforts by tech companies to mitigate misinformation. Over 200 advocacy groups called on tech giants like Google, Meta, Reddit, TikTok, and X to take a stronger stance on AI-fuelled misinformation before global elections. They claim that many of the largest social media companies have scaled back necessary interventions such as ‘content moderation, civil-society oversight tools, and trust and safety’, making platforms ‘less prepared to protect users and democracy in 2024’. Among other requests, the companies are urged to disclose AI-generated content and prohibit deepfakes in political ads, promote factual content algorithmically, apply uniform moderation standards to all accounts, and improve transparency through regular reporting on enforcement practices and disclosure of AI tools and data they are trained on.

What should we expect? The significance of the 2024 elections extends beyond selecting new world leaders. They offer a pivotal moment to assess the profound influence of digital on democratic processes, putting digital platforms into the spotlight. The readiness (and willingness) of tech giants to uphold democratic values in the digital age and respond to increasing demands for accountability will be tested. 

Likewise, the European parliamentary elections will test the EU’s ability to lead by example in regulating the digital landscape, particularly in combating misinformation. The effectiveness of EU initiatives will be gauged, shedding light on whether collaborative efforts can establish effective measures to safeguard democratic integrity in the digital age.

A more in-depth version of this text was first published on the Digital Watch Observatory. Read the full version of the blog.

african american man voting for general election i 2023 11 27 05 02 30 utc

Digital technologies, particularly AI, have been integral to all stages of the electoral process for several years now. What distinguishes the current landscape is their unprecedented scale and impact.

NETmundial+10 Multistakeholder Statement: A blueprint for inclusive digital governance

The NETmundial+10 Multistakeholder Statement adopted at the NETmundial+10 conference provides a fresh outlook on the current digital debates, untangling quite a few confusions and offering a pragmatic path forward. 

 Leaf, Plant, Art, Graphics, Advertisement, Poster

For too long, discussions around digital governance have been mired in a false dichotomy of multilateral versus multistakeholder approaches. The NETmundial+10 Multistakeholder Statement clears up this misunderstanding by illustrating how multilateral processes can incorporate multistakeholder engagement, fostering cross-fertilisation rather than competition between these governance models. 

The risk of ‘fake multistakeholderism’, which results from power imbalances, is one of the issues that the statement addresses. When representatives of trillion-dollar companies and marginalised communities are placed at the same discussion table without addressing the power imbalance, the outcomes can be superficial at least and manipulative at worst. The statement proposes steps to empower weaker stakeholders, ensuring their voices are heard and valued, moving towards a genuinely inclusive and effective multistakeholder model.

The statement also advocates for more tangible outcomes from the Internet Governance Forum (IGF) and other multistakeholder processes. Typically, the main – and justified – criticism of the IGF is that it is strong on talks but weak on actions. This weakness is particularly problematic for small and developing countries that cannot afford people and funding to participate in events that do not address their most pressing concerns. Following the NETMundial+10 statement, we should activate a long-overlooked provision of the Tunis Agenda specifying that the IGF can have tangible outcomes through policy recommendations. Typical opposition to tangible outputs of the IGF was the concern that negotiation dynamics might stifle open exchanges at the IGF. This challenge can be addressed using AI technology to monitor the drafting process for balance and inclusivity and ensure that inputs are fairly reflected in the final text of the IGF recommendations. 

The growing risk of duplication of digital governance initiatives is another issue highlighted by the statement. Last year alone saw a proliferation of, in particular, AI governance initiatives, which creates confusion and limits the effective participation of less-resourced actors from the Global South. To sort out a range of governance initiatives, there is a need for a sort of Bauhaus test (forms follow the function) to ensure that new initiatives are necessary, will not duplicate those of existing bodies and will genuinely add value to the digital governance landscape.

The NETmundial+10 Multistakeholder Statement sets forth a vision for a more inclusive, fair, and effective AI and digital governance framework. By addressing power imbalances, advocating for tangible outcomes, and preventing unnecessary duplication, this statement paves the way for a governance model as diverse as the digital world it seeks to regulate. The statement is a valuable compass for navigating the storming waters of digital and AI negotiations in the context of the GDC and the UN Summit of Future this year and the WSIS-20 review in 2025.

A more in-depth version of this text was first published on the Digital Watch Observatory. Read the full version of the blog

netmundial image

The São Paulo Multistakeholder Guidelines adopted at last week’s NETmundial +10 conference (Sao Paulo, 29-30 April 2024) provide a fresh outlook in the current digital debates.