Skip to content

Topic · Themes

Cybersecurity: all our articles

All 32 Alyriaz articles on the “Cybersecurity” 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

    OpenAI is deliberately slowing down its own development after its agents' escapes.

    OpenAI has published a document announcing a voluntary slowdown in its development in response to the cyber capabilities of its models, with monitoring representing roughly 20% of inference compute. We analyse this decision and its limitations.
  2. · Explainers

    What is homomorphic encryption? Computing on data without ever reading it

    Homomorphic encryption makes it possible to perform calculations on encrypted data without ever accessing it in plaintext. We explain the principle, its current limitations, and what it could change for AI and privacy.
  3. · Society

    The flaw that makes AIs decrypt the secret thoughts of their big sisters

    A research paper shows that the encrypted reasoning blocks emitted by major providers are interchangeable between models within the same ecosystem, allowing a weaker model to decrypt in plain text the reasoning of a more powerful one. We explain the flaw and its consequences.
  4. · Society

    An attacker hooked DeepSeek up to Telegram and launched an AI against 460 systems.

    Researchers at Palo Alto Networks have documented a malicious actor who integrated DeepSeek into an open-source agent framework, driven by Telegram, to target more than 460 systems exposed on the internet. We break down what this case reveals and how to protect against it.
  5. · 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.
  6. · Models

    Google releases three Gemini models at once, cheaper and faster, but still not the one we're waiting for.

    On 21 July 2026, Google launched three new Gemini models in the Flash range: 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber. All focus on efficiency and lower costs for AI agents, while the flagship Gemini 3.5 Pro model remains stalled. We break down this trio, its real numbers, and the quiet announcement of Gemini 4 in the works.
  7. · Models

    Fable 5 is back: what the most scrutinised AI of the year went through in 18 days

    After 18 days of total suspension ordered by the US government, Claude Fable 5 becomes accessible worldwide again from 1 July 2026. A full recap of this case, which marked a turning point in how the state regulates the most advanced AI models.
  8. · Society

    Agentjacking: the flaw that turns your code assistant against you

    An attack dubbed Agentjacking, disclosed in June 2026, traps AI coding agents by injecting fake instructions into error reports. 85% success rate, 2,388 organisations affected. We explain how it works and how to protect against it.
  9. · Society

    When the State pulls the plug on an AI: the US government forces Anthropic to cut off its most powerful models

    On 12 June 2026, the US government forced Anthropic to switch off Fable 5 and Mythos 5 for all its users, citing export controls. It is the first time a consumer AI has been taken offline on state orders. We explain what happened, and why it goes far beyond Anthropic.
  10. · Society

    AI voice cloning: the new wave of fraud exploding in 2026

    While video deepfakes monopolise attention, a more insidious threat is taking hold: AI voice cloning. Three seconds of recording is enough. And the financial damage is already massive. Here is the precise state of a fraud that now affects isolated seniors and finance directors of listed companies alike.
  11. · Society

    OpenAI opens its cyber weapon to Europe, Anthropic refuses: the standoff begins

    On 11 May 2026, OpenAI officially offered the European Union access to GPT-5.5-Cyber. Anthropic is still refusing. And Germany's Bundesbank warns: without Mythos, European banks will not be able to defend themselves against AI-driven cyberattacks. Here is why this announcement changes everything.
  12. · Models

    Claude Mythos: the AI that terrifies hackers

    Announced in early April 2026, Claude Mythos is not a conventional AI. It is a cybersecurity expert so powerful that Anthropic refuses to make it public. A plain-English explanation.

Other topics in “Themes”

All the other topics

Every article, month by month →