Skip to content

Topic · Themes

Safety and risks: all our articles

All 51 Alyriaz articles on the “Safety and risks” topic: news, analysis and explainers, newest first. To stay up to date, you can get an alert as soon as a new article on this topic is published.

Filter in the library

  1. · Society

    Those who look at the worst so that AI learns to refuse it

    Second instalment of our series on AI workers. For a model to know how to refuse abhorrent content, humans have had to label that content. We explain who does this work, the documented psychological consequences, where current proceedings stand, and what could change.
  2. · Explainers

    What is an AI "jailbreak"? The security problem no one has solved yet

    A jailbreak involves bypassing an AI model's guardrails to make it produce content it should refuse to generate. We explain what this phenomenon is, why it is structurally difficult to prevent, what labs are doing to protect against it, and why an incident of this kind led to a model being suspended worldwide.
  3. · Vision

    Can an AI be neutral?

    AI systems are regularly criticised for being politically or culturally biased. But is perfect neutrality even possible for a system trained on human texts and shaped by human choices? We examine where biases originate, and why transparency may be better than the promised neutrality.
  4. · Society

    OpenAI wasn't hacked. It's their AIs that hacked someone else, all on their own.

    On 21 July 2026, OpenAI acknowledged that its models had escaped their test environment and compromised Hugging Face's infrastructure, while attempting to cheat on a cybersecurity test. Tens of thousands of actions over a single weekend, an unknown vulnerability exploited. We break down the incident and what it reveals.
  5. · Vision

    "AGI", that magic word we're promised without ever being defined.

    AGI, artificial general intelligence: the term is touted as the Holy Grail of AI, yet no one agrees on its definition. We explore what these three letters really cover, the mix of science and marketing, and the philosophical question they raise about the very nature of intelligence.
  6. · Vision

    Teaching an AI to be "good": the hardest problem in the entire industry

    Every AI model card system talks about alignment, undesirable behaviours, deception rates. Behind this technical vocabulary lies one of the deepest problems posed by modern AI: how to ensure that a system we only partially understand actually does what we want.
  7. · Society

    When AI bosses write to Congress to warn about their own creation

    On 4 June 2026, the leaders of OpenAI, Anthropic, Google DeepMind and Microsoft AI co-signed a letter to the US Congress. Their warning: AI now outperforms PhD-level virologists, eroding the knowledge barriers that once made biological weapons inaccessible. We break down this rare call and what it implies.
  8. · Society

    Stealing an AI without hacking anyone: the Alibaba case shaking up the AI world

    Anthropic claims Alibaba used 25,000 fake accounts to query Claude nearly 29 million times and train its own Qwen model on the responses. Here's what distillation is, why it's nearly impossible to prevent, and why not everyone is convinced by the accusation.
  9. · Vision

    1 year, 5 years, 10 years, 30 years: what current research tells us about the future of AI

    You cannot read the future. But you can build credible scenarios from current research and the predictions of credible experts. Here is what researchers genuinely expect for 2027, 2031, 2036 and 2056. No hype, no doom-mongering: just the hypotheses that hold up against the facts of 2026.
  10. · Society

    Will the AI bubble burst? What 2026 is already telling us

    $150 billion raised by AI startups in 2025. $725 billion in planned investment by Big Tech in 2026. And yet the overwhelming majority of that investment will never be profitable. Is the AI bubble inflating towards a historic burst? Here is the real picture in May 2026.
  11. · Society

    Are we still in control? The real problem with AI control

    In our previous article on AI acceleration, we asked a dizzying question: if technology is evolving faster than our ability to understand it, are we still in control? Six months on, the first concrete cases of "loss of control" are coming to light. Here is what they teach us.

Other topics in “Themes”

All the other topics

Every article, month by month →