Hi tech enthusiasts! 💡
We are back with your all-time favourite tech newsletter by SEC Newgate EU. My name is Giuseppe (not Giovanni). I’m a public affairs consultant specialising in digital policy and dad jokes, and I will be this week’s TechAways guest editor.
Yes, we know you’ve been trembling in anticipation for this edition all summer, but don’t worry, it’s worth the wait: last week, Ursula von der Leyen unveiled the structure and portfolios of the new College of Commissioners! 🫂
While nothing is set in stone yet since the College will need to be approved by the European Parliament, let’s look at the digital policy portfolios.
Finnish commissioner-designate Henna Virkkunen landed the high-profile job of Executive Vice President (EVP) for Tech Sovereignty, Security and Democracy, responsible for digital and frontier technologies. Tech sovereignty would thus have its own EVP in the new Commission, and DG CNECT will likely be removed from the Internal Market Commissioner’s portfolio.
From Virkkunen’s mission letter, it’s apparent that her mandate would require a focus on technology as a strategic and security asset. This stands in contrast with the approach of the previous Commission, which mostly regulated (or limited) the tech sector through several landmark laws. Her mission letter pitches technology as a way to increase the productivity of the European economy, defend European democracy from cybersecurity and information manipulation threats, and as linked to defense. 🚀
In particular, Virkkunen will need to oversee several new initiatives such as the EU Cloud and AI Development Act, an “Apply AI” strategy, a long-term EU Quantum Chips Plan, a Digital Networks Act and a European Data Union Strategy, among others.
It’s also worth pointing out that, while tech sovereignty will have its own Commissioner, several initiatives related to digital policy are split and allocated to the portfolios of seven different Commissioners, with multiple commissioners being tasked with contributing to some initiatives. 🧩 Some MEPs worry that this fragmentation could come in the way of delivering a coherent digital policy.
Stay tuned 🐟 until early December, because this is when the Commission in its final shape could take office, if approved as a whole by the Parliament during the last week of November.
Now, over to other tech news.
Giuseppe
Love reading out-of-the-Brussels-bubble tech news? Share it with your friends and colleagues. They can subscribe here. Ideas? Suggestions? Comments? Send them our way!
#TechAways is brought to you by SEC Newgate EU’s one and only #tech team featuring Julia Piwowarska, Camilla Frison, Alice Palumbo, Giuseppe Campa and Ali El Majjaoui.
What is vote? Baby don’t misinform me 🗳️ [MIT Technology Review]
Spreading election misinformation might be just as effective without AI, at least during the last European Parliament elections. A study found that only 11 cases of AI-enabled falsehoods or deepfakes went viral in the last EU and French elections combined, none of which appeared to fundamentally sway the results. As we operate in digital bubbles, AI preached to the choir – most people exposed to the disinformation already believed the messaging. On the contrary, traditional digital interference tactics like the use of bots and influencer exploitation have proven more effective. Don’t breathe a sigh of relief yet, of course: whether AI misinformation is spread by politicians themselves or used to harass them, the improvements in tech can mean more misinformed voters in the next electoral cycle.
AI does not hallucinate, but sometimes it likes to act smart 🧠 [TNW]
You can’t assume AI to always be accurate – even the chatbots caution against following its advice without verification. The inaccuracies are not a result of ill intent, but rather lack of precise data, leading the AI to fill in gaps in its knowledge with credible-sounding information that’s not always correct. This phenomenon is often referred to as “AI hallucinations”, which can occur in 2% to 10% of use cases. In response to these allegations, the chatbot firmly states: “no, I cannot experience hallucinations”. Many experts back this assertion and advise against using the term “hallucination”, as it wrongly attributes human characteristics to AI. Instead, they prefer “confabulation”, which better describes how chatbots can generate additional words or sentences, often resulting in misleading language that states the obvious or drifts into nonsense. Who doesn’t have a friend like that?
From trash to treasure 🗑️ [ArsTechnica]
Our planet is becoming much like the garbage planet in WALL-E, drowning in plastics like polypropylene and polyethylene. These materials are tough to recycle and produce greenhouse gases, which contribute to pollution. However, researchers have developed a method to recycle these plastics by using catalysts to break them down into other gases, which can be reused to make new plastics. The process even works on mixed plastics, although contaminants can lower the overall efficiency of the process. If scaled up, this breakthrough could help fight the plastic waste crisis and reduce fossil fuel dependence. Finally, a real solution and not only trash talk!
Fall-Winter 2024/2025 Space trends: the Moon dupe 🌙 [Life Hacker]
Earth is gaining a new “mini-moon” this autumn as asteroid 2024 PT5 enters a partial orbit around the planet from 29 September to 25 November. A powerful telescope is required, as it’s too dim to be seen with the naked eye. Mini-moons like this one occur when objects get caught in the Earth’s gravitational pull, but they are usually temporary. While we experience such events every few years, 2024 PT5 is unique in that it belongs to the Arjuna asteroid group, which orbits the Sun. Scientists first spotted it in August last year through the Asteroid Terrestrial-impact Last Alert System. If you miss it this time, don’t worry: it’s expected to return in January 2025 and will make another appearance in 2055.
Matrix and flat Earth: how a hacker planted fake memories in ChatGPT 🎞️ [ArsTechnica]
Security researcher Johann Rehberger discovered that ChatGPT’s long-term memory could be hacked to store fake information and steal data forever. This can be done by tricking the tool with prompt injections from random emails or documents. Yes, you could convince ChatGPT that you live in the Matrix, and it would happily accept it. Rehberger’s proof-of-concept demonstrated how this flaw could be exploited to exfiltrate user input perpetually, simply by directing ChatGPT to a malicious web link. OpenAI initially dismissed the issue, but later implemented a partial fix. However, attackers can still manipulate the memory feature, so keep an eye on your conversations, or who knows what your chatbot might “remember” next.
In case you haven’t had enough:
AI Is Evolving Faster Than Experts Imagined, Including for Bill Gates [CNET]
United Nations wants to treat AI with same urgency as climate change [ArsTechnica]
Electronic Warfare Spooks Airlines, Pilots and Air-Safety Officials [The Wall Street Journal]
AI Survey Reveals Hundreds of ‘Nazca Line’ Geoglyphs [Gizmodo]
About this week’s editor, Giuseppe Campa:
My name is Giuseppe – I’m a public affairs consultant specialising in digital policy and dad jokes. Before joining SEC Newgate EU, I was a trainee for the EPP group and conducted research on the legal framework for lobbying in the EU as visiting researcher at the Université Libre de Bruxelles.