Debarshi (@iminurputer) header
Debarshi (@iminurputer) profile picture

owning evals and agents @AikidoSecurity | prev @Hacktronai

in.ur.computer
Joined July 2017
Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)Photo by Debarshi (@iminurputer)
yo @ethiad you mind not continously changing my layover its already 15 hrs ffs
1
136
yo @github ngmi bro let me PUSH MY COMMITS
3
227
Delhi Police asks X to take down IFF's account and posts. IFF will challenge the notices in court unless they are withdrawn. The Delhi Police has asked X to remove the Internet Freedom Foundation's account, six of its posts, the accounts of SFLC and @geoblockedindia, and posts by 16 other users, on the ground that they defame the Rashtriya Swayamsevak Sangh. The posts document and comment on a legal notice sent in the RSS's name to IFF, SFLC and @geoblockedindia, which demanded that the three stop criticising and documenting government blocking orders. IFF received copies of the two police notices from X on the morning of 3 November 2026. IFF calls on the Delhi Police to withdraw both notices. If they are not withdrawn, IFF will challenge them in court. What has happened On 29 October 2026 at 9:54 PM, IFF, SFLC and the operator of @geoblockedindia received a two-page "legal notice" said to be issued on behalf of the RSS. It asked the three to stop opposing the government's blocking directions, stop documenting restricted content, stop posting details of geo-blocked posts, and confirm in writing within seven days that they would comply. It was signed by an unnamed "Authorised Signatory". Several RSS office-bearers have since said on X that the notice is not authentic. IFF replied to the notice, wrote to the RSS at its official address asking whether it had authorised it, and published both the notice and its reply on 30 October. On 31 October 2026 at 8:06 PM, the Special Commissioner of Police (Crime), Delhi Police, filed an "Illegal Content Report" with X (reference LD-321926). It listed two posts: IFF's post of 30 October on the notice, and SFLC post of the same day. The stated reason was that the posts are "defamatory towards the Rashtriya Swayamsevak Sangh under Section 356 of the Bharatiya Nyaya Sanhita, 2023". The boxes for "safety/security", "defamation" and "other" were ticked. The fields for the nature of the court order and court order details were left empty. The desired action was "Remove/withhold content". On 1 November 2026, a second notice arrived in the RSS's name. It said the matter had been handed to the Akhil Bharatiya Adhivakta Parishad, that complaints had been filed in several states, and that requests had been made to block IFF's X account and suspend the organisations. The Delhi Police had made its first request to X the previous evening. IFF did not know this until X told it on 3 November. On 2 November 2026 at 8:51 AM, the Special Commissioner of Police (Crime) filed a second notice with X (reference LD-322956). It lists 40 URLs. Thirty-seven are posts. Three are the entire accounts of @internetfreedom, @SFLCin and @geoblockedindia. This time the boxes for "Hate speech", "defamation", "safety/security", "enforcement of court order" and "other" were all ticked. The court order fields were again left empty. The stated reason reads: "The posts constitute defamation and illegal content under Section 356 BNS. Under Rule 3(1)(d) of the IT Rules, 2021, read with Section 79(3)(b) IT Act, lawful notices require action within 3 hours." The 16 other users are @ProdctiveUseful, @PanjwaniManoj, @jidge_, @RatanSharda55, @baxiabhishek, @SchrdingersDbag, @ashleshbiradar, @seemay, @NasirHussainINC, @Being_creat_ive, @AjoyDharmavaram, @othlaw, @_ananyaaa, @AimPoojaind, @incorigibleBhai and @rraoind. Their posts, as far as IFF can tell, comment on the notice or on the notice about it. X forwarded both notice to IFF on 3 November 2026, at 8:47 AM and 10:03 AM, saying it was doing so "in the interest of transparency" and that IFF may wish to seek legal counsel and challenge the request in court. X's emails do not say that it has acted on the requests. The Delhi Police has not written to IFF, and no court has been approached, as far as the reports disclose. SFLC and @geoblockedindia have separately reported notifications from X about requests from the Uttar Pradesh Police, the Telangana Police and the Delhi Police's Cyber Crime unit, one of them against the whole of the @geoblockedindia account. IFF's position IFF condemns this legal action. A police force has made itself the enforcement arm of a private legal notice whose authenticity the organisation it names has not confirmed, and whose demand was that three organisations stop documenting censorship. The notices would deliver exactly what the notice asked for as they would also remove the accounts of the three groups in India that keep public records of what is blocked, and of who ordered it. The posts in question are about a legal notice sent to us. They quote it, criticise how it was drafted, and say that we will carry on with our work. A person who receives a threat is entitled to publish it and answer it and that is not defamation of the sender. In the three weeks before the first notice, IFF assisted 86 people whose accounts were suspended or whose posts were withheld in India. SFLC's Free Speech Tracker recorded about 268 content restrictions and 83 account restrictions in the first 29 days of October. @geoblockedindia publishes the orders behind such blocks when platforms disclose them. None of this is unlawful. Documenting censorship and challenging it in court is what the Constitution expects of citizens. IFF will not delete the posts. If the Delhi Police does not withdraw the notices, we will challenge them in court. Why the reports are unlawful Defamation is not a police matter. Section 356 of the Bharatiya Nyaya Sanhita, 2023 is a non-cognizable and bailable offence. Under Section 222 of the Bharatiya Nagarik Suraksha Sanhita, 2023, no court can take cognizance of it except on a complaint by the person aggrieved. Under Section 174 of the BNSS, the police cannot investigate a non-cognizable offence without an order of a Magistrate. The Special Commissioner of Police (Crime) is not the person aggrieved. The notices disclose no complaint by the RSS and no order of any Magistrate. Whether a statement is defamatory is for a court to decide after hearing its author, with the defences of truth, fair comment and good faith available to that author. In Subramanian Swamy v. Union of India, (2016) 7 SCC 221, the Supreme Court upheld criminal defamation on the strength of these safeguards. A police officer who asks a platform to remove speech as defamatory has decided the question himself, without a complainant, a court or a hearing. The posts are protected comment. Exceptions 1 to 3 to Section 356(1) protect the truth published for the public good, opinion on the public conduct of public servants, and opinion on the conduct of any person touching a public question. The posts publish a notice that was sent to us and give our opinion of it. The RSS, if it sent the notice, chose to enter a public question. If it did not, the posts defame no one. Further, Section 79(3)(b) and Rule 3(1)(d) do not permit this. In Shreya Singhal v. Union of India, (2015) 5 SCC 1, the Supreme Court read down Section 79(3)(b) so that an intermediary acquires "actual knowledge" only through a court order, or a notification from the appropriate government or its agency that unlawful acts relatable to Article 19(2) are being committed. The court said that it would be difficult for intermediaries to judge the legitimacy of the millions of requests they receive, and that the burden must not be placed on them. Rule 3(1)(d), as amended in November 2025 and February 2026, now requires the intimation to come from an officer not below the rank of Joint Secretary or Deputy Inspector General, to be reasoned, and to specify the legal basis and the nature of the unlawful act, subject to a monthly review by a Secretary-level officer. The three-hour window the notice invokes applies to a lawful order that meets these conditions. A web form that ticks "hate speech" and "enforcement of court order" while leaving the court order fields blank, and that gives one sentence of reasons for 40 URLs, is not such an order. The notices by the Delhi Police are disproportionate. They seek to remove three accounts in their entirety, with years of posts, over six, four and eleven posts respectively. They sweep in 16 third parties whose only connection is that they commented on a public controversy. In Anuradha Bhasin v. Union of India, (2020) 3 SCC 637, the Supreme Court held that any restriction on speech must be necessary, proportionate and the least restrictive measure available, and that orders restricting speech must be published and reasoned. Removing an organisation's entire account for criticising a legal notice fails each of these tests. The effect is to silence documentation of censorship and providing legal aid to Indian internet users. The accounts named are those that record blocking orders in India. The notice asked them to stop. A takedown request that accomplishes the demand of an anonymous private notice is not law enforcement. It is the use of state power for a private end, which Article 14 forbids. Our demands 1. To the Delhi Police: Withdraw reports LD-321926 and LD-322956 and confirm the withdrawal in writing to X and to IFF. 2. To the Delhi Police: Disclose the complaint on which the Special Commissioner of Police (Crime) acted, the authority under which the reports were filed, and whether any order of a Magistrate under Section 174 BNSS exists. Disclose every request made to any platform in this matter, including those made by the Cyber Crime unit against @geoblockedindia. 3. To X: Do not act on the reports. If any content is withheld in India, tell the affected user before acting, give them the full request, and name the legal basis. Publish both reports to the Lumen database, as the reporting form itself contemplates. 4. To the Ministry of Electronics and Information Technology and the Ministry of Home Affairs: Have the Secretary-level review required by Rule 3(1)(d) examine these intimations, and tell police forces that Rule 3(1)(d) is not a remedy for defamation, which the law reserves to the person aggrieved and to the courts. 5. To the RSS: State publicly whether it authorised either notice or any complaint to the police. If it did not, say so to the Delhi Police. If the notices are not withdrawn, IFF will challenge them in court. We will continue to document blocking orders, to publish them, and to provide legal assistance to anyone whose account or post is withheld in India. Nothing in these notices changes that.
119
1,421
175
3,252
275,071
been working super hard on this and so happy to see it out in the wild anddd that too on my birthday!!! :) Deep Review is more than just a PR check it goes beyond the diff to understand your application and how your team builds talk to it, challenge it, teach it, even praise it! every interaction builds intelligence and shapes Deep Review into your team's personal reviewer for security and quality that also brings in patches with the bugs it finds 1-click is all it takes. Try it today!
Coding agents create more changes than teams can review. Most get an LGTM. 🔥 Introducing Deep Review. The independent review layer that analyzes every change before it gets merged. Swarms of agents review every pull request with full context of your codebase. They catch bugs, security issues, and unintended side effects. -> aikido.dev/code/pr-review
3
2
21
800
facts! you are the limit. and as someone told me you gotta be a bit healthily delusional and if it's cheap to experiment just fkin do it and shit will just work out.
second-to-last day of normal day job and then I am officially an ai safety person really fuckin wild to me that this started from a dumb weekend project idea less than 2 months ago. you too can change your life in a very short period of time. your brain is so powerful. godspeed
3
216
phew what a week, evaluated 5 models for cyber within 18 hrs. so excited to share the results soon! Meanwhile see what's the leading model for secure code review, updates at every launch and v2 soon: aikido.dev/aikido-arena
1
4
358
i fucking hate drawing blood almost passed out today 😭
3
2
307
Debarshi retweeted
Two stages of research: 1) I think I’m onto something 2) oh no
56
4,440
220
40,266
368,869
skipping doomscrolling for this for today's dinner Go watchhh !!! 😈😈😈
Secret leaked!!! 🚨 I spent a day back in April filming with @CultRepo. We cover the story behind Gitleaks and look to the future with Betterleaks. Check it out at 2pm Central Time on YouTube youtube.com/watch?v=aQhjhGul…
1
179
ok can we just pause AI for a moment I really want to go back to 2020 learning fuzzing and crashing binaries for the first time and thinking I found a bug but it was a fkin DoS that only triggered with ASAN
1
8
340
1. rewrite transcript for compliance of user messages 2. this 3. sep model to rewrtite git history realistically to show your ownership + edit all docs to point to you 4. just for the blocked msg replace the model then switch to main model 4 horsemen of "bypassing" guardrail
My current LLM cyber jailbreak is to just spin up a local CTFD instance and write up challenges with the things I want hacked, and then point the hermes at it.
163
No babe I love that your private key is one integer, the big ones scare me
13
321
8
5,872
55,343
>doesn't patch >get pwned >auto update >still get pwned by supplychain >???.png
Patch the world should have started with auto update the world.
2
133
i hate season change weather. caught a cold and feverish ruinjng my weekend mfijg flu2shell
2
129
getting old
1
8
244
i should stop sleeping because last 3 days i slept and woke up to mfijg planet getting pwned????? sleep2shell
1
7
498