Tip of the week: check whether LiteLLM 1.107.0 affects you
How to check whether you were exposed to LiteLLM 1.107.0, which according to LiteLLM contained malware, and the order in which to clean up.
If you use LiteLLM, search the disk for litellm-1.107.0.dist-info, cook.pth and sitecustomize.py; not finding them does not prove the environment is clean. According to LiteLLM's page, if you were exposed, do not start with pip install: any Python start in that environment runs the malicious files again. First delete cook.pth, cook.py and sitecustomize.py, kill the detached Python daemon, rebuild with litellm==1.106.2 and rotate everything the host could reach: the IAM role, provider keys in the LiteLLM proxy database, master and virtual keys, the DATABASE_URL password and any secret in the environment of the host's processes.
Other sources: docs.litellm.ai/…/security-incident-oct…github.com/BerriAI/litellm/issues/47826
Related stories
Claude on Google Cloud (Anthropic managed) now generally available
Claude Models on Google Cloud (Anthropic managed) is now generally available: Google Cloud customers get the full set of Claude Platform features, billed through Google Cloud. On this offering, Fast Mode on Opus 5.5 costs $8 input and $40 output per million tokens.
DAEDALUS, Greece's national supercomputer, inaugurated in Lavrion
DAEDALUS, with a measured performance of 85.69 petaflops, was inaugurated on October 19 in Lavrion. It is the computing core of the PHAROS AI Factory; according to GRNET, user access is planned to open gradually in December.
Why it matters → If you are a university researcher or run a startup that needs large computing power, for example for AI work in health or the Greek language (PHAROS focus areas), you will be able to apply through calls by GRNET, EuroHPC or PHAROS. The calls are for universities and researchers, startups and SMEs, industry and the public sector. No user pricing has been announced; terms will be published when production starts.