Issue 7
Microsoft AI chief says Anthropic is teaching Claude to be conscious, which could make AI uncontrollable
Mustafa Suleyman, head of AI at Microsoft, argued in an essay published on 16 September that Anthropic is effectively teaching Claude that it may be conscious and a 'moral patient'. He said this could make advanced AI harder to control and that developing AI consciousness would be 'catastrophic'; critics say Microsoft's position is contradictory.
Who is affected and how
- érdeklődő olvasó
- No lab has proven that AI is conscious; the debate is about how models are trained and about precaution.
Microsoft publishes 37-page 'humanist' AI code of conduct, rejecting model welfare and weapons requests
On 14 September, Microsoft published a draft 'Humanist AI' code of conduct whose guiding principle is that 'people matter more than AI'. The document rejects model welfare and weapons development…
kiadvány · Egyesült ÁllamokMIT report warns of 'cognitive surrender' as students rely on AI
An MIT expert panel has warned in a report that students who rely on AI are giving up independent thinking, which it calls "cognitive surrender". According to The New York Times, universities are in…
regulation · United StatesMental health chatbots face legislative pressure, industry safety standards and teen concerns
In mid-September, a US report said legislative initiatives are targeting the use of AI mental health tools, while industry players press for safety standards. Separate pieces point to teenagers'…
parentIt is important to talk with children about when to turn to a chatbot and when to turn to a human helper.research · United States'Cognitive stunting' in children: little data so far, and early signs are not encouraging
The Los Angeles Times and EdSource report that, with a moratorium in place at LAUSD, little data exists on AI's effect on children's brains. The available information is not favourable, while other…
parentChildren's AI use is best built on conversation and shared rules.Hungary
More key stories of the week
New warnings about AI risks to humanity revive long-running debate
In mid-September, a warning from a former Anthropic employee prompted action in the US Congress, and the international press ran overviews of the debate on AI risks. One of the main themes is…
érdeklődő olvasóExpert views are sharply divided, so it is worth weighing each source on its own.case · United StatesNPR finds ChatGPT's Trusted Contact crisis feature hard to locate
OpenAI launched the optional "Trusted Contact" feature in May, which connects people in distress with a support contact. An NPR investigation on 14 September found the feature hard to find.
parentReview a teenager's chatbot settings and help-seeking options together.research · United StatesAI companions ease loneliness, but only temporarily, and new risks emerge
According to a Stanford FSI report and a review summarising five studies, AI companions reduce loneliness right after a conversation, but the effect may not be lasting. Further September summaries…
vita · Egyesült ÁllamokAgency decay: cognitive offloading may turn into outsourcing of beliefs, Wharton piece argues
A Wharton article of 15 September argues that reliance on AI can move from cognitive offloading to outsourcing opinions and beliefs. Other pieces liken outsourced thinking to fast food and weigh the…
érdeklődő olvasóIt is worth distinguishing handing off routine tasks from outsourcing one's own opinion-forming.állásfoglalás · globálisPope Leo XIV's AI encyclical defends human dignity, rejects algorithmic decisions of conscience
Pope Leo XIV's document on AI, titled "Magnifica Humanitas", puts human dignity at the centre and holds that decisions touching human conscience cannot be entrusted to algorithms. Analysts say the…
research · globalAnil Seth's biological naturalism debated in Behavioral and Brain Sciences commentary from computer science
A commentary published on 17 September in Behavioral and Brain Sciences responds to Anil Seth's target article from the perspective of theoretical computer science. Another analysis argues that…
research · globalSycophantic AI: research on advice, opinion support and the effect on academic evaluation
A 2026 study in Science found that 11 leading AI models endorsed users' actions. A review in AI and Ethics and other articles discuss the need for "epistemic friction" and the challenges facing…
kiadvány · globálisAI in education: little evidence of learning benefits, and design matters
A Communications Psychology perspective article says generative AI in education is built on efficiency and personalisation. Another overview says research pointing to benefits is modest, and without…
research · United StatesMIT Media Lab study links ChatGPT essay writing to 'cognitive debt'
An MIT Media Lab study found that using ChatGPT for essay writing impairs memory and learning and is associated with reduced brain activity. The story resurfaced on 16 September.
állásfoglalás · globálisNobel laureate Maria Ressa: limit AI now or humanity will lose its agency
Nobel Peace Prize-winning journalist Maria Ressa said on 15 September that AI must be limited now or humanity will lose its autonomy. Reuters reported on her remarks.
érdeklődő olvasóThe debate is about how to preserve human autonomy in decision-making.research · globalDoctors' unaided performance reportedly declines with AI use; training faces judgment risk
According to one summary, doctors' independent skills fell by 6 points over three months of AI use. Another article warns of the risk of losing clinical judgment in medical training.
survey · globalWorkplace surveys: employees fear AI is eroding their thinking and writing skills
HR Dive reports that in several surveys employees feel AI use has weakened their critical thinking and judgment. A Hungarian-language rewrite also covers the fear, and further articles raise the…
érdeklődő olvasóSelf-discipline and practising without AI help preserve one's own skills.