[afnog] African ai infrastructure

[sm+afrinic at elandsys.com]

Hello,
At 08:57 AM 14-06-2026, Loganaden Velvindron wrote:
>It is sad that one ai model is currently being restricted due to 
>national security reasons in the usa.

There was an academic paper about "boundary point jailbreaking of 
black-box LLMs" [1].  It includes some information about bypassing 
the restrictions built into a LLM.

The possibility that Anthropic Fable's restrictions could be bypassed 
caused security concerns in the U.S.  That is why foreigners are no 
longer allowed to access that AI model.

The capabilities of an AI model are often highly exaggerated [2].  A 
few open source software projects have been complaining about code 
submissions [3] generated by AI.  There was a significant increase in 
vulnerability reports.  Some of those reports are probably AI-generated [4].

Regards,
S. Moonesamy

1. https://arxiv.org/pdf/2602.15001
2. 
https://www.reuters.com/business/fears-unfettered-hacking-spurred-by-anthropics-mythos-ai-model-overstated-2026-05-20/
3. https://devtalk.blender.org/t/ai-contributions-policy-proposal/44202
4. Liu, D., Chang, O., metzman, J., Sablotny, M., & Maruseac, M. 
(2024). OSS-Fuzz-Gen: Automated Fuzz Target Generation