AI models cross a new capability threshold
Barely a month after its launch, Claude Mythos Preview is already improving its performance. According to the latest tests published by the UK’s AI Security Institute (AISI), Anthropic’s model now beats its own results from the first evaluations and also outperforms OpenAI’s GPT-5.5 on several cybersecurity exercises. This particularly rapid progress is reigniting the debate on whether security mechanisms can keep pace with the development of the most advanced artificial intelligence models.1
Unprecedented performance in cyber tests
The AISI regularly evaluates artificial intelligence models through scenarios that replicate real cybersecurity operations. In this new round of testing, Mythos Preview became the first model to complete certain ranges previously considered out of reach, notably the Cooling Tower scenario, which had never before been solved by an AI. The researchers point out that this improvement came without a new official generation of the model, demonstrating that AI capabilities continue to advance between versions thanks to optimisations made by their developers.1
Capabilities are advancing faster than expected
The results observed exceed the projections made a few months earlier by the British researchers. In February 2026, the AISI already estimated that AI models’ ability to carry out cybersecurity tasks was doubling roughly every five months. The arrival of Mythos Preview and GPT-5.5 has accelerated this trend. Experts now believe that models are advancing faster than the evaluation scenarios themselves, making it increasingly difficult to measure their true limits.1
A new reality for cybersecurity
Improved model performance brings obvious benefits for defence teams. AI can help detect vulnerabilities, analyse code or speed up the patching of security flaws. But these same capabilities could also be exploited by malicious actors to identify weaknesses in IT systems more quickly. Specialists therefore believe that every technological advance must now be accompanied by new control, evaluation and oversight mechanisms to limit the risk of misuse.2
Benchmarks are reaching their own limits
The AISI researchers also acknowledge that their evaluation methods are starting to show their limits. Current scenarios impose a ceiling of 2.5 million tokens in order to compare models fairly. Yet the latest AI systems appear capable of going well beyond this limit. The researchers believe that larger compute budgets could improve their performance further, suggesting that the real capabilities of the most advanced models are still partly underestimated by current evaluation tools.1
Cybersecurity must evolve at the same pace as artificial intelligence
This development is a reminder of why Cybersecurity is becoming more strategic than ever. Professionals no longer just have to protect infrastructure against traditional attacks; they must also understand the capabilities of new generations of artificial intelligence, anticipate how they will be used and put in place mechanisms able to govern their deployment. Continuous model evaluation, risk management and behaviour monitoring are now becoming essential components of digital security.
A race that now goes beyond performance
The progress of Mythos Preview shows that competition between artificial intelligence players is no longer solely about model performance. It also concerns the ability to guarantee their safety, control and responsible use. As these systems become more autonomous and effective, cybersecurity is becoming an inseparable part of their development and a strategic issue for the entire digital ecosystem.
360° IT News — decoding technology to understand the jobs of tomorrow.
Further reading
The rapid progress of Mythos Preview illustrates how quickly the capabilities of the most advanced artificial intelligence models are accelerating. On a related topic, read our article “Claude Fable 5 leaves the free plan: Anthropic bets on professional use”, which analyses how Anthropic is now repositioning its most powerful models towards businesses and high-value use cases.
References
1. UK AI Security Institute. (2026). Cyber capability evaluations of Claude Mythos Preview.
https://www.aisi.gov.uk
2. Anthropic. (2026). Project Glasswing and Frontier Model Safety.
https://www.anthropic.com
GitHub Copilot bulks up for business: coding agents come under IT control
Copilot is no longer just there to suggest the next line GitHub Copilot is continuing its transformation. Originally known mainly for completing code in the editor, the tool is now evolving towards agents capable of exploring a project, modifying files, using tools...
PDFs, presentations, audio: Adobe wants to turn Acrobat into a content factory
Acrobat no longer just wants to open your documents For more than thirty years, Acrobat has mainly been used to read, create, edit, sign and share PDFs. Adobe now wants to give it a much broader role. On 9 September 2026, the company announced new capabilities based...
Salesforce plugs Claude into its CRM: Claudeforce prepares for life after the interface
What if the CRM soon no longer needed to be opened? For decades, using a CRM meant navigating between customer records, dashboards, menus and forms. With Claudeforce, Salesforce and Anthropic want to change this logic: rather than forcing users to go into Salesforce...
Kimi K3, Mistral, DeepSeek: Oracle turns OCI into an AI model hub
Oracle wants to make OCI a one-stop shop for AI Oracle is stepping up its efforts in the AI cloud battle by significantly expanding the catalogue of models available in OCI Enterprise AI. On 4 September 2026, the group announced in particular the integration of Kimi...
Agentic Data Cloud: Google rethinks data for the age of AI agents
AI agents are forcing Google to rethink data architecture AI agents promise to search for information, make decisions and act directly within company systems, but this autonomy depends on one essential condition: accessing the right data and, above all, understanding...
Google Cloud tackles the hidden bill of AI agents
AI agents raise a new problem: how much does a task really cost? As companies move from chatbots to AI agents able to work through multiple steps, the question of cost becomes much more complex. A chatbot generally handles a request and generates an answer, whereas an...
Workday, Navan, Avalara: ChatGPT Work moves into business software
ChatGPT Work is gradually leaving the chat window With GPT-6 Astra, OpenAI is no longer simply trying to improve the quality of ChatGPT’s answers. The company wants to bring its AI closer to the applications in which employees actually work. With the arrival of GPT-6...
From chip to cloud: Groq raises 350 million to scale up
Groq now wants to sell cloud rather than just chips Groq is pursuing a major strategic transformation. Long seen as one of Nvidia’s challengers thanks to its LPUs (Language Processing Units) specialised in artificial intelligence model inference, the group is now...
Muse Code: Meta launches its AI agent to tackle giant codebases
Meta wants to bring AI agents into complex software projects Meta is stepping up its efforts in AI-assisted software development with Muse Code, a new agent designed to work on large codebases, where traditional assistants can quickly lose the context needed to...
Qwen3.8-Max reportedly coded for 16 days without human intervention
Alibaba tests how far an autonomous AI developer can go Alibaba claims to have reached a new milestone in agentic software development with Qwen3.8-Max, its new flagship model launched in early August 2026. In an internal test, the model reportedly worked fully...
