I am Vinoth, from San Diego. I currently work on Datacenter HW security at OpenAI. I used to lead the silicon security architecture and silicon security operations teams at #Google. Before that, I worked on mobile silicon security at #Qualcomm.
Much of my professional work so far has been about improving the security of #Pixel and other #Android devices. I will mostly talk about #mobilesecurity, #androidsecurity and #infosec.
I sometimes invest in seed stage start-ups, primarily technology companies. My portfolio includes Modumate, Akido Labs, kia.ai, Zendoc and Zeoauto. If you are building something cool, hit me up.
| https://twitter.com/salt___doll | |
| https://www.linkedin.com/in/vinoth-deivasigamani-04b7a92/ |
For user privacy in Cloud AI, TEE is a huge step up. I think that's the right direction. E2E doesn't make sense here as you don't trust one of the ends (i.e. Cloud).
However, adverse interests will use the above fact to lower the bar for broader privacy from E2E to TEE. This would include use cases and experiences where TEE a significant erosion of privacy, such as human-to-human messaging.
F.B.I. Arrests C.I.A. Official With $40 Million in Gold Bars in His Home
"From last November to March, the court papers say, Mr. Rush asked for, and received, “a significant quantity of foreign currency and tens of millions of dollars in gold bars for work-related expenses.”
Doesn't the CIA require receipts for reimbursements?
https://www.nytimes.com/2026/05/27/us/politics/fbi-arrest-cia-official-gold-bars.html
Erin Brockovich has a website that raises awareness about concerns around AI data centers. It even has a map of community-reported data centers.
Ironically, the website looks like it is AI generated/assisted.
Codex is truly becoming the everything app.
I barely use ChatGPT anymore. I use Codex for coding, business book keeping, tax returns prep, searching email, searching slack, terminal work, repo analysis, reading, journaling, and honestly a huge chunk of my daily workflow now.
Codex is truly becoming the everything app.
I barely use ChatGPT anymore. I use Codex for coding, searching email, searching slack, terminal work, repo analysis, reading, journaling, and honestly a huge chunk of my daily workflow now.
I find it interesting that the cyber LLMs like GPT-5.4-Cyber* and Mythos** corroborate this. They are not particularly trained to identify vulnerabilities. They are (more or less) fine-tuned versions of general purpose models that remove the cyber refusals. In other words, they are simply excellent developers learning security on the job.
* "..we are fine-tuning our models specifically to enable defensive cybersecurity use cases, starting today with a variant of GPT‑5.4 trained to be cyber-permissive: GPT‑5.4‑Cyber." https://openai.com/index/scaling-trusted-access-for-cyber-defense/
** "We did not explicitly train Mythos Preview to have these capabilities. Rather, they emerged as a downstream consequence of general improvements in code, reasoning, and autonomy. " https://red.anthropic.com/2026/mythos-preview/
#cybersecurity #hiring #mythos #gpt54cyber #openai #anthropic
To be an effective security engineer, you need both systems expertise and security expertise. However, for most complex systems (Eg: Mobile, servers, PCs), it is a lot easier and quicker to train a systems expert in security than a security expert in systems. There are some exceptions, of course, such as cryptanalysis and formal methods. You will know when you have an exception case.
#cybersecurity #hiring #mythos #gpt54cyber #openai #anthropic
A security hiring strategy that I preach and practice: If the choice is between an (a) expert in the system we are securing with little to no security expertise and (b) a security expert with little to no system expertise - choose (a).
Codex has been open source since the beginning. Turns out no one has cloned it and destroyed OpenAI's moat. Just saying.
