Shlomi Hod

🚨 Submission deadline is approaching for the Responsible Synthetic Data (RSD) Workshop @ AAAI 2026 📢 The RSD workshop at AAAI 2026 (27th January, 🇸🇬 Singapore) focuses on responsible practices for synthetic data with/for foundation models.  🌐 Website: responsible-synthetic-data.github.io

The new Claude Tag feature seems extremely useful, but at the same time, a dangerous bargain for enterprises because of the pricing model and the risk of lock-in. The four big changes together mean that you interact with Claude as a coworker instead of a tool (the same Claude instance for everyone instead of each worker; soaks up tacit knowledge without your telling it; acts on its own; and does so asynchronously). All clearly very useful, but completely flips the interaction paradigm. anthropic.com/news/introduci… Let’s talk about lock-in. As far as I can tell, Claude maintains its own memories in this new way of working; the human team members can’t see and edit them. (System administrators presumably can, but they have other things to do!) Tacit knowledge thus goes from a weakness of AI agents to a major strength — it seems inevitable that as teams and orgs start to use Claude this way, it will become the main queryable repository of all their tacit knowledge, creating dependence and stickiness. Effectively, Claude is a coworker that you can’t fire without *every* team losing workflows and know-how. By the way, it also seems to introduce a new and pervasive security risk, since Claude can be integrated into private channels as well, and can be given access to repositories and tools even if the users in that channel don’t have access to them. Anthropic has introduced an interesting but complicated access control model to handle all this: claude.com/blog/agent-ide… But I’m not sure I trust people to understand and implement it correctly, nor the LLMs to be sufficiently robust against threats like prompt injection. What about pricing? Claude is not like regular coworkers, because it bills for every token it produces. And it can do an unbounded amount of work, asynchronously and without being asked. In the current model, when AI is a tool, enterprises set per-user budgets, which creates accountability and keeps cost somewhat manageable. When everyone shares a Claude, it will be much harder to track and control spending. Of course you can set a token budget, but turning off Claude for the month for everybody when the budget is hit risks bringing work to a screeching halt. When AI companies talk about the next stage of AI being a “drop-in replacement” for human workers, it should be understood not as a technical innovation but a business model innovation, enabling more value capture and rent extraction. AI companies are no longer competing for a share of enterprises’ IT budgets but rather a share of their entire labor spend, which is orders of magnitude bigger. Claude Tag is a big milestone in this evolution. This shift is very good for AI companies, but it is unclear if it is good for their customers.

Honoured that our 2016 paper, Robust Estimators in High Dimensions without the Computational Intractability, w/ Ilias Diakonikolas, Daniel Kane, Jerry Li, Ankur Moitra, Alistair Stewart, was awarded the 2026 Gödel Prize This is the highest award for papers in theoretical CS. 1/7

We discovered that it's possible to create an AI-driven computer worm using an open-weight model that anyone can download. This work was conducted in a lab walled off from the outside world, & shared only after removing details that could aid bad actors. nytimes.com/2026/06/02/tec…

the beauty of a jagged frontier is that every failure is the user's fault and every breakthrough is the model's glory

AVERI just published an analysis of audit-related legislation in the US. We survey the current landscape, discuss challenges with audit requirements and pathways for addressing them, and make our first endorsement of specific legislation.

"AI" is not a stochastic parrot.🦜 I wrote this piece a couple weeks ago, but it was hard for me to finish up given AI's role in society and war over the past few weeks. I should share it at some point though. Not perfect, but here it is. @margarmitchell/no-ai-is-not-a-stochastic-parrot-a99e57766bed" target="_blank" rel="nofollow noopener">medium.com/@margarmitchel…

@jmduke @random_walker It was just recently increased by 1! termswatch.io#61a0b264

One of the funnier GDPR disclosures I've seen in a while.

"Professionals and litigants should be aware that their AI conversations may be discoverable."

Today, @evaluatingevals is introducing Every Eval Ever, a unified, open data format and public dataset for AI evaluation results.

Today we’re releasing the International AI Safety Report 2026: the most comprehensive evidence-based assessment of AI capabilities, emerging risks, and safety measures to date. 🧵 (1/17)

Frontier models are starting to display a shift in capabilities in offensive security. Over the past few weeks, we are seeing growing evidence of a change: publicly available frontier models are now reliably solving complex, well-defined offensive-security tasks.

📢 In case you missed it: the first-cycle deadline for FORC 2026 is *tomorrow*, November 11. Submit your best work on mathematical research in computation and society, writ large. Too soon? We'll also have a second-cycle deadline on February 17, 2026. CfP link below!👇

Two out of the four reasons they give here are bizarre science fiction relating to "model welfare" - I'm sorry, but I can't take seriously the idea that Claude 3 Opus has "morally relevant preferences" with respect to no longer having its weights served in production

1/ NEW: We propose a new black-box attack on LLMs that needs only text (no logits, no extra models). It's generic: we can craft adversarial examples, prompt injections, and jailbreaks using the model itself👇 How? Just ask the model for optimization advice! 🎯

Privacy in LLMs is not just Memorization! We reviewed 1322 papers (2016–25) across ML, NLP & SEC: 92% fixate on memorization/chat leaks. We map 5 urgent problems + a roadmap, to prevent surveillance, inference, aggregation and other negative outcomes.

📌 Key Topics Include: - Lifecycle Uses & LLM-Driven Generation - Safety & Robustness - Privacy, Security & Data Governance - Fairness, Bias & Representation - Explainability, Interpretability & Uncertainty - Metrics & Tooling for Trustworthy Use - Critical Perspectives

Foundation models increasingly leverage synthetic data for training while simultaneously generating synthetic datasets for downstream applications. This workshop centers on the responsible development and use of synthetic data with and for foundation models

🚨 Submission deadline is approaching for the Responsible Synthetic Data (RSD) Workshop @ AAAI 2026 📢 The RSD workshop at AAAI 2026 (27th January, 🇸🇬 Singapore) focuses on responsible practices for synthetic data with/for foundation models.  🌐 Website: responsible-synthetic-data.github.io

We spent the last year evaluating agents for HAL. My biggest learning: We live in the Windows 95 era of agent evaluation.