LiteLLM 1.107.0 compromised: CERT-In ki warning aur kya check karein

23 October 2026 ko CERT-In, India ki national agency for cyber security incidents, ne compromised LiteLLM package par advisory CIAD-2026-0053 jaari ki. Developers LiteLLM ko AI chatbots aur automations ke andar use karte hain, taaki wo ChatGPT, Gemini, Claude aur doosre models se jud sakein, aur version 1.107.0 me password stealer hone ki report aayi hai. Ise aap khud install nahi karte, lekin aapke liye kaam karne wala developer, agency ya AI vendor shayad kar chuka ho. Ye update CERT-In ki advisory, project ke public GitHub issue aur PyPI page par based hai, jo teeno 24 October 2026 ko padhe gaye.

What happened: LiteLLM 1.107.0 was compromised

• CERT-In kehta hai ki version 1.107.0, jo PyPI (Python Package Index) se baanta gaya, "reportedly" ek targeted software supply-chain compromise ka shikaar hua, aur GitHub Container Registry tatha Docker Hub ke Docker images bhi reportedly compromised the.
• Project ka GitHub issue number 47826 kehta hai ki 1.107.0 ki files "contain a credential stealer and a remote-code loader", jo files banne se pehle jodha gaya aur public source tag me nahi hai. Dono sources ek lookalike repository Berri-AI/litellm ki taraf ishara karte hain, project ki asli repository ki taraf nahi.
• CERT-In ke hisaab se ye code Python start hote hi chal sakta hai, chahe koi application LiteLLM import kare ya na kare, isliye usi environment ke doosre Python programs bhi exposed ho sakte hain. Issue jodta hai ki stealing wala part Windows par bhi chalta hai.
• CERT-In ke mutabik ye browser me saved passwords aur session cookies, SSH aur GPG keys, AWS, Azure aur Google Cloud ke logins, environment variables me rakhe secrets, aur Cursor aur GitHub Copilot jaise tools ke tokens jama karne ki koshish karta hai. LiteLLM proxy par wo wahan rakhi AI provider keys tak bhi pahunch sakta hai. Doosra part aur code download karke chala sakta hai.
• Issue ki 21 October 2026 ki timeline: 13:23 UTC tak PyPI ne 1.107.0 hata diya tha, aur 17:13 UTC tak Docker tags latest, main-latest aur main-stable ab bad image par point nahi karte the. 24 October 2026 ko PyPI 1.107.0 ko "quarantined" aur 1.106.2 ko current release dikhata hai. Issue 1.106.2 ko clean batata hai aur use pin karne ki salah deta hai; CERT-In pehle verify karne ko kehta hai.

What is not known: kitne systems par bad version install hua, attacker build ya publishing process me kaise ghusa (issue ke authors kehte hain ki unhe nahi pata), aur kaun se AI tools ke andar LiteLLM use hota hai.

Why it matters for a small business

Risk ye nahi hai ki ChatGPT ya Gemini galat answers dein. Risk us computer ya server ka hai jahan bad version chala: wahan saved passwords, cloud logins aur AI provider keys copy ho gayi ho sakti hain. Hamare hisaab se chori hui AI key koi aur use kar sakta hai aur bill aapke paas aata hai, aur chori hua cloud login aapki website ya customer data tak ka darwaza khol sakta hai. Hamari guide customer data privacy and AI tools dekhiye.

Agar aap ChatGPT, Gemini ya Claude sirf browser ya phone app me use karte hain, to aap LiteLLM install nahi karte. CERT-In ki advisory developers, DevOps aur cloud teams ke liye hai. Chinta un logon ki hai jo aapke liye AI banate aur chalate hain.

Who should care

• Wo owners jinke developer, agency ya freelancer ne AI chatbot, assistant ya automation banaya hai.
• Wo teams jo kai AI providers ki keys sambhalne ke liye apna AI gateway ya proxy chalate hain.
• Wo businesses jo AI software khareedte hain: vendor ke software ke andar LiteLLM ho sakta hai aur bahar se aap use dekh nahi sakte. Hamari guide what to check before you pay for an AI tool me aur questions hain.

What changes for you

What to do this week

1. Un sabhi AI chatbots, assistants aur automations ki list banaiye jo aapke business ke liye bane ya chal rahe hain, aur har ek ko sambhalne wale person ya company ka naam likhiye.
2. Har ek ko neeche wala message bhejiye. Writing me "we do not use LiteLLM" ek achha answer hai.
3. Agar answer haan ho, to unse CERT-In ke steps follow karne ko kahiye: har wo system dhoondhna jisme 1.107.0 install hua, clean image se rebuild karna aur keys "a trusted, clean environment" se badalna, jaise CERT-In kehta hai.
4. Apne AI provider accounts ke usage aur billing pages dekhiye ki koi aisa use to nahi jo aap pehchante nahi, aur provider de to spending limit set kar dijiye.

Hello [Name],

CERT-In has published advisory CIAD-2026-0053 about LiteLLM version 1.107.0 (a PyPI package and some Docker images), released on 21 October 2026. Please reply in writing:

1. Do the tools, chatbots or automations you built or run for us use LiteLLM, directly or inside another tool?
2. If yes, did any computer, server or container ever install or run version 1.107.0, or pull the Docker images latest, main-latest or main-stable?
3. If yes, which passwords, API keys and cloud logins could those systems reach, and when will they be changed?

Thank you.

Ye general information hai, security audit nahi. Agar aapko lagta hai ki koi system affected tha, to kisi qualified security professional ko involve karein. Ek rule bhi madad karta hai ki koi bhi work computer par bina poochhe naye AI tools install na kare: dekhiye simple AI rules for your team.

What to watch next

• Kya LiteLLM team batati hai ki build kaise compromise hua, jo issue ke hisaab se abhi pata nahi hai.
• Kya CERT-In advisory CIAD-2026-0053 ko naye details ya aur affected versions ke saath update karta hai.
• Kya jin agencies aur AI vendors ko aap use karte hain, wo aapko batate hain ki wo affected the ya nahi.

Not sure where AI can help your business safely? The free AI Business Checkup takes about three minutes.