TLV Partners' 2026 TLV AI 50 names 30 Israeli AI researchers and 20 founders across OpenAI, DeepMind, Meta, NVIDIA, Weizmann and Technion, plus 7 flagged papers including Lightricks' LTX-2.
TLV Partners' 2026 TLV AI 50 names 30 Israeli AI researchers and 20 founders working inside OpenAI, DeepMind, Meta, NVIDIA, Weizmann Institute of Science, and Technion, alongside seven papers published in the past year with at least one Israeli author. The list shows a split pattern: 14 of the 30 named researchers still hold a position at an Israeli university, while most of the 20 named founders build companies inside Israel, including eight who remain in stealth mode.
The TLV AI 50 is an annual list published by TLV Partners, a Tel Aviv venture capital firm, naming 50 Israeli individuals working across artificial intelligence research and company building. The 2026 edition covers 30 researchers, 20 founders, and 7 flagged papers from the past year with at least one Israeli co-author. TLV Partners describes the list as a snapshot of a community rather than a ranking, and separately announced a planned grant program for Israeli AI researchers called TLV AI Scholar.
How many researchers, founders, and papers made the 2026 list?
The 2026 TLV AI 50 splits into three named groups totaling 50 people, plus a separate set of 7 flagged papers. The table below breaks down each group with representative names pulled directly from the list.
| Category | Count | Representative names |
|---|---|---|
| Researchers | 30 | Boaz Barak (Harvard, OpenAI), Yoav Goldberg (Bar Ilan University, Allen Institute for AI), Michal Irani (Weizmann Institute of Science) |
| Founders | 20 | Amnon Shashua and Shai Shalev-Shwartz (AAI, Mobileye, AI21), Omri Geller (Run:ai, acquired by NVIDIA) |
| Papers | 7 | LTX-2 (Lightricks), From Tokens to Thoughts (Chen Shani, Ravid Shwartz Ziv, with Dan Jurafsky and Yann LeCun) |
| Total people named | 50 | 30 researchers plus 20 founders |
Where do Israel's AI researchers actually work?
Fourteen of the 30 named researchers hold a current position at an Israeli university or research institute, most concentrated at Weizmann Institute of Science, Technion, and Tel Aviv University. The remaining 16 work exclusively at foreign universities or foreign AI labs, including Harvard, Princeton, Oxford, Columbia, OpenAI, DeepMind, and Meta.
| Institution type | Named organizations | Example researcher |
|---|---|---|
| Israeli academia | Weizmann Institute of Science, Technion, Tel Aviv University, Bar Ilan University | Daniel Soudry (Technion, also NVIDIA) |
| Foreign academia | Harvard, Princeton, Oxford, Columbia, Cornell Tech, UC Berkeley, NYU | Boaz Barak (Harvard, also OpenAI) |
| Foreign AI labs | OpenAI, DeepMind, Meta, Thinking Machines Lab, Google Gemini | Yossi Gandelsman (OpenAI) |
| Chipmakers and vision labs | NVIDIA, Black Forest Labs | Hila Chefer (Black Forest Labs) |
Three of the 30 researchers hold a role at NVIDIA specifically: Daniel Soudry, Gal Chechik, and Shie Mannor, all three also teaching at an Israeli university at the same time. That overlap, an Israeli professorship paired with a chipmaker research title, is the most common dual affiliation on the list.
Why does TLV Partners say Israel leads in AI security and vision?
TLV Partners identifies AI security and computer vision as the two fields where Israeli AI work sits at the technical frontier, based on the concentration of named researchers and founders in each category on the 2026 list. The firm distinguishes the current AI-security cohort from an older generation of cybersecurity companies that sold defensive tools to corporate security officers, describing the newer companies as research-led and standards-setting for how large companies deploy AI safely, rather than sales-led.
On computer vision, TLV Partners points to Israeli academic labs that rank among the most cited in the field, plus Israelis holding technical leadership roles inside the vision programs at major labs. Michal Irani at Weizmann Institute of Science and Hila Chefer at Black Forest Labs are two named examples of that vision concentration, alongside Tamar Rott Shaham, also at Weizmann, whose work spans vision-language interpretability.
Which paper on the list is drawing the most attention?
LTX-2, an open-source audio-visual foundation model built by Lightricks, is the most technically significant paper on the 2026 list. The model pairs a 14-billion-parameter video stream with a 5-billion-parameter audio stream inside one dual-stream transformer, generating synchronized video and audio from a single prompt instead of adding sound in a separate pass. Lightricks published the full paper on arXiv under the identifier 2601.03233 and released the model with open weights, a departure from the closed releases most large labs use for comparable video-generation systems.
The other six flagged papers cover narrower research problems, most from teams working at Technion, Weizmann, and Hebrew University of Jerusalem. The table below lists all seven with their core finding in one line each.
| Paper | Lead authors | Core finding |
|---|---|---|
| LTX-2 | Yoav HaCohen, Zeev Farbman (Lightricks) | Open-weights audio-video foundation model matching proprietary systems at lower compute cost |
| From Tokens to Thoughts | Chen Shani, Ravid Shwartz Ziv, with Dan Jurafsky and Yann LeCun | Large language models match human concept boundaries but compress away fine-grained nuance |
| BetaDescribe | Yonatan Belinkov's Technion group | A model generates functional protein descriptions for proteins with no annotated homologs |
| Induction Meets Biology | Gal Pomerants, Yaniv Nikankin, Yonatan Belinkov, Ora Schueler-Furman | Protein language models detect sequence repeats using induction heads plus biology-specific neurons |
| Dual Mechanisms of Spatial Variable Binding | Tamar Rott Shaham, Shoval Messica | Vision-language models bind objects to spatial relations mainly inside the vision encoder |
| LLMs Generate Harmful Responses Using a Distinct Mechanism | Hadas Orgad, Yonatan Belinkov | A small, sparse set of parameters drives harmful outputs across multiple harm categories |
| Agents of Chaos | Natalie Shapira, Hadas Orgad, Tamar Rott Shaham | Two-week red-teaming of autonomous agents found identity spoofing and false success reporting |
Are Israel's AI founders building at home or exiting abroad?
Most of the 20 Israeli AI founders on the 2026 TLV AI 50 are building companies inside Israel rather than relocating to join foreign labs, and eight of the 20, Ariel Shaulov, Dan Barzilay, Ido Ben Shaul, Lavi Shapiro, Seri Khoury, Uri Rolls, Yaron Singer, and Yossi Adi, are working in stealth mode without a public company name attached yet. The founder list also includes veterans of prior Israeli exits building second companies rather than leaving the country: Amnon Shashua co-founded Mobileye before founding AI21, and Omri Geller founded Run:ai before its acquisition by NVIDIA.
This pattern, where founders from one exit start a second company inside Israel rather than relocating, shows up elsewhere in Israeli tech. The four founders of Wiz took a similar path after Google's acquisition of the company, moving into new ventures and capital deployment rather than leaving the country.
The founder cohort also shows NVIDIA acting as a repeat acquirer of Israeli AI infrastructure companies. Beyond Run:ai, NVIDIA's Israeli holdings include the chip and networking businesses covered in Israel's AI silicon stack, spanning Annapurna Labs, Habana Labs, and Mellanox. The second-act pattern among founders is documented in more detail in the Wiz founders' post-acquisition paths.
Who are the most recognizable names on the 2026 TLV AI 50?
Eight names on the list carry the widest public recognition, spanning both the researcher and founder groups. Each entry below answers who they are and where they currently work.
| Name | Current role | Why they are on the list |
|---|---|---|
| Amnon Shashua | Co-founder, AI21; co-founder and CEO, Mobileye | Founded Mobileye, then co-founded AI21 Labs and AAI with Shai Shalev-Shwartz, making him the list's clearest example of a repeat Israeli AI founder |
| Boaz Barak | Professor at Harvard University; Member of Technical Staff, OpenAI | Holds a joint Harvard-OpenAI role, one of the clearest examples of a researcher who left an Israeli institution for a foreign lab |
| Yoav Goldberg | Professor and Research Director, Bar Ilan University; Allen Institute for AI | Combines an Israeli professorship with a research-director post at a major US-based nonprofit AI lab |
| Michal Irani | Professor, Weizmann Institute of Science | Named as one of the researchers anchoring Israel's frontier position in computer vision |
| Shafi Goldwasser | Professor, UC Berkeley | Turing Award-winning cryptographer whose inclusion signals the AI-security cohort draws on Israel's older cryptography and theory-of-computation base |
| Omri Geller | Co-founder and CEO, Run:ai (acquired by NVIDIA) | Founded the GPU-orchestration company NVIDIA acquired, one of the list's named examples of an AI-infrastructure exit |
| Hila Chefer | Research Scientist, Black Forest Labs | Named alongside Michal Irani as part of Israel's concentration in computer-vision research, working at the lab behind the FLUX image models |
| Yair Carmon | Senior Lecturer, Tel Aviv University; researcher, Safe Superintelligence | Combines an Israeli academic post with a role at Safe Superintelligence, the AI-safety lab co-founded by former OpenAI chief scientist Ilya Sutskever |
What does the 2026 TLV AI 50 show about Israel's AI position?
Israel's contribution to the Israeli AI companies and infrastructure map splits along two tracks that TLV Partners' own data makes visible: 16 of 30 researchers train and publish exclusively at foreign universities or labs, while most of the 20 founders build and stay inside Israel, with AI security and computer vision as the two fields where both tracks concentrate. TLV Partners plans to fund more of the research side directly through a new program, TLV AI Scholar, aimed at Israeli AI researchers, with details to follow.
CONCLUSION
The 2026 TLV AI 50 maps a country whose AI researchers mostly work abroad and whose AI founders mostly build at home, concentrated in security and vision. The list is a snapshot, not a ranking, and TLV Partners has not yet said when it will publish the next edition.
