20 AI Startups That Launched in New York This Month (June 2026)
New York City closed the first half of 2026 as the de facto global launchpad for agentic AI. Between May 19 and June 18, at least twenty AI-native companies surfaced through stealth exits, public debuts, or NY Tech Week appearances with verifiable New York ties. The cohort spans LLM infrastructure, agent identity security, verified consumer research, construction intelligence, and conversational commerce — and it has already pulled more than $198 million in disclosed seed and Series A capital into the five boroughs. This editorial maps the wave for operators, buyers, and investors who need neutral, source-backed context: thirteen of the companies carry fully verified Professional Business Directory listings, and seven additional names are tracked from NY Tech Week press coverage pending full profiles.
Why New York, why now
NYC Tech Week (June 1–7, 2026) compressed founder and investor attention around agentic systems at the exact moment enterprises began shifting from copilot pilots to production-grade AI infrastructure. Three structural advantages explain why launches clustered in New York even when engineering teams remain distributed:
- Buyer density. Financial services, global media, luxury retail, and commercial real estate — the sectors most exposed to AI-driven workflow change — are headquartered within a 20-mile radius of Midtown. When Minerva needed a launch customer to stress-test its Agentic Data Scientist against real first-party fan data, the NBA’s league office was a subway ride from its Brooklyn office.
- IP and rights stakeholders. Publishers, record labels, and sports leagues concentrated in NYC are the natural counterparties for training-data licensing products like FanClub AI and Taste Labs. The city houses the legal and business-affairs teams that negotiated the $1.5 billion Bartz v. Anthropic settlement — a number that framed every rights conversation during Tech Week.
- Capital proximity. Sequoia, a16z, Union Square Ventures, and a fresh crop of AI-specialist funds maintain partner-level presence in the city, shortening the path from demo to term sheet. Lightspeed Venture Partners, which backed A Security with a $32 million follow-on, had partners on the ground for the entirety of Tech Week.
Funding snapshot: $198M in six weeks
Across the twenty launches tracked, disclosed funding totals roughly $198.6 million, with a clear concentration in security and infrastructure plays. Security-focused startups (Offroad, A Security, Tenet Security) alone account for approximately $50 million — $7 million for Offroad, $37 million across two rounds for A Security, and $6 million for Tenet — underscoring how quickly “agent risk” has become a standalone budget line for enterprise buyers. Consumer research and marketing (Pogo at $32 million total raised, Minerva at $20 million) account for another $52 million, while LLM infrastructure (Concentrate AI at $5.1 million, plus undisclosed rounds for Tiger Data and Thread AI at a reported $20 million Series A) pulls in roughly $25 million. Taste Labs, which emerged from stealth on June 16 with an $18.5 million seed co-led by CRV and Amplify Partners, represents the largest single round among training-data and IP plays.
| Segment | Companies | Disclosed capital |
|---|---|---|
| Agent & identity security | Offroad, A Security, Tenet Security | ~$50.0M |
| LLM infrastructure & data | Tiger Data, Concentrate AI, Thread AI | ~$25.1M |
| Consumer research & marketing | Pogo, Minerva, Knotch, Voicelyt | ~$52.0M |
| Training data & IP | FanClub AI, Taste Labs | ~$18.5M (Taste Labs disclosed) |
| Vertical AI (CRE, construction) | YieldStack, Trunk Tools | Undisclosed |
| Robotics & commerce | Mecka AI, Kernso, Shoppable | ~$60.0M (Mecka, multi-year) |
Thirteen startups with verified directory profiles
Each row below links to a structured Professional Business Directory listing carrying ten or more cited sources, map coordinates (city-level where street addresses remain unpublished), and explicit editorial disclaimers. Funding figures are drawn from dated press releases unless otherwise noted.
| Company | Category | Focus | Public milestone (2026) | Directory |
|---|---|---|---|---|
| FanClub AI | IP / Training data | IP trust layer & Library Vault for ethical AI training — female-led by Deborah Harpur, a 25-year veteran of entertainment licensing for LVMH, Netflix, Warner Bros., and Paramount. Offices in NY, LA, London, expanding to Seoul. Pre-revenue and currently fundraising against the backdrop of 125+ active infringement lawsuits and the $1.5B Bartz v. Anthropic settlement. | May 27 — Business Wire launch | PBD listing |
| Offroad | Security | Agentic identity security: autonomous agents that investigate, govern, remediate, and verify access risks across human users, machine identities, OAuth applications, service accounts, and AI agents. Audited 2,890 public OAuth apps and found that roughly one in three — representing more than 1.85 billion installs — carried serious structural security concerns. Published the free OhAuth security catalog alongside its launch. | Jun 4 — stealth exit + $7M seed | PBD listing |
| A Security | Security | Autonomous offensive security platform that deploys offensive and defensive AI agents simultaneously to discover, validate, and eliminate real-world exploit paths before attackers can use them. Raised $37M total across two rounds ($5M seed led by Cyberstarts, $32M follow-on from Lightspeed and Cyberstarts), with angel checks from Wiz CEO Assaf Rappaport and Cyera CEO Yotam Segev. Catalyst: Anthropic’s Mythos model exposed thousands of zero-days, rendering traditional pentesting obsolete. | Jun 8 — stealth exit + $37M | PBD listing |
| Pogo | Consumer research | Purchase-verified consumer panel built from a consumer app with 3M+ opted-in U.S. users. Tracks $470B in transaction volume and has visibility into 1 in every 150 U.S. shopping trips. Platform launches AI-moderated video interviews at scale, reaching verified buyers of specific SKUs in hours. Customers include Hershey, Nestlé Purina, Chomps, Funko, Bain, and The Trade Desk. Founded in 2020 by Oskar Melking, Shikhar Mohan, and Dom Wong. | Jun 9 — platform launch + $32M total raised | PBD listing |
| Minerva | Marketing | Brooklyn-based agentic consumer marketing platform. Raised $20M from The General Partnership, 8VC, Lingotto Innovation, Topology Ventures, and NBA Investments. Built two workstreams on GPT-5.5 in collaboration with OpenAI: an Agentic Data Engineer that standardizes fragmented first-party data, and an Agentic Data Scientist that generates predictive models from natural-language prompts. Claims 3.4x ROAS improvement on paid media. Customers include the NBA, Luxury Presence, and Wander. | Jun 9 — public launch + $20M (Brooklyn) | PBD listing |
| Tiger Data | Infrastructure | Ghost — a purpose-built Postgres service for AI agents, created by the team behind TimescaleDB. Delivers disposable, ephemeral database environments that agents can spin up and tear down without touching production infrastructure. | Jun 9 — general availability | PBD listing |
| Concentrate AI | Infrastructure | Managed LLM gateway offering access to 130+ frontier and open-source models through a single API with zero token markup. Founded by Ari Jacoby (CEO) and Todd Lieberman (CTO), who previously built Deduce (acquired by CHEQ). Backed by True Ventures and RRE Ventures with a $5.1M pre-seed. Free tier available with tokens at cost. | Jun 10 — stealth exit + $5.1M pre-seed | PBD listing |
| Knotch | Content infrastructure | ACE — an AI experience layer for the conversational web, built on Google Cloud and Vertex AI. Designed to help brands own measurement and optimization across AI-mediated search and assistant surfaces. | Jun 16 — launch with Google Cloud / Vertex AI | PBD listing |
| Voicelyt | Consumer / coaching | AI voice coach with a proprietary Voice Score vocal analysis engine. Public launch positioning it at the intersection of executive coaching, speech therapy, and productivity tools. | Jun 16 — public launch | PBD listing |
| Tenet Security | Security | Runtime protection for autonomous AI agents. Patent-pending “Agent-side Simulation” predicts an agent’s next actions before execution and can block risky behavior in real time. Founded by former Cisco AI Defense builders Barak Sternberg and Nevo Poran. Tenet Threat Labs coined the term “Agentjacking” — malicious instructions embedded in emails, logs, or databases that hijack agent behavior — and validated the technique across 100+ enterprise environments. Raised $6M seed led by The Westly Group and MizMaa Ventures. | Jun 17 — stealth exit + $6M seed | PBD listing |
| Taste Labs | Training data | Post-training evaluation data company founded by Thais Castello Branco, former growth lead at Exa. Thesis: standard RLHF methods degrade aesthetic quality and manufacture the uniformity the industry now calls “AI slop.” Sells preference datasets, reasoning data, rubrics, and reinforcement-learning evaluation environments to frontier model labs. Team of 20 drawn from Exa, Palantir, and Mercado Libre. First domain: design. $18.5M seed co-led by CRV and Amplify Partners. | Jun 16 — stealth exit + $18.5M seed | PBD listing |
| YieldStack | Vertical / CRE | AI-native commercial mortgage lender matching platform. Matches borrowers against 60+ deal attributes across 180+ lender programs for loans ranging from $50,000 to $1 billion. Charges 50–100 basis points at closing versus the industry-standard 1–2%. Founded in 2025 by CEO Daniel Chesney, who brought experience from real estate investment banking and private credit. | Jun 17 — public platform launch | PBD listing |
| Trunk Tools | Vertical / construction | Cortex — an AI intelligence layer for construction drawings, trained on 10M+ construction documents over four years with major U.S. general contractors. Powers a suite of workflow agents (TrunkReview, TrunkRFI, TrunkBid, TrunkSubmittal, TrunkRegister, TrunkText) that compress submittal review from two hours to under ten minutes. Deployed on hundreds of active U.S. jobsites. Customers include Gilbane, Suffolk, HITT, DPR, Harkins, Consigli, Torcon, McGough, and Cleveland Construction. SOC 2 Type II certified. Founded in 2021 by CEO Dr. Sarah Buchner. | Jun 17 — Cortex + workflow agents | PBD listing |
Seven more names from NY Tech Week (profiles pending)
These companies meet our editorial “launch activity” criteria but do not yet carry dedicated PBD places pages. We list them with primary signals only — no endorsement, no NAP claim.
| Company | Focus | Signal | Reference |
|---|---|---|---|
| Thread AI | Lemma composable AI orchestration (ex-Palantir founders) | Jun 2026 Series A $20M — 131 Varick St, NYC | Official site |
| Kernso | Agentic commerce resolution engine | Kernso Corporation — HQ New York, founded 2026 | Official site |
| Shoppable | Universal Checkout MCP for AI assistants — 500M+ SKU catalog, founded by Heather Udo, protected by 4 U.S. patents (filed 2011). Named one of Entrepreneur Magazine’s “100 Brilliant Companies.” | May 12, 2026 MCP launch — New York. Early data shows AI-referred visitors converting at multiples of traditional search. Positioned as the neutral storefront for the multi-assistant era versus Amazon’s walled garden. | Official site |
| Mecka AI | Human-motion data for robot training at scale | NYC HQ — $60M raised (2025–2026 trade coverage) | Official site |
| Neural Defend | Deepfake detection for BFSI authenticity | 1st place — Everything AI pitch, NY Tech Week Jun 3 | — |
| Tangering AI | Last-mile delivery dispute evidence automation | 2nd place — Everything AI pitch, NY Tech Week Jun 3 | — |
| Corseco | Verification layer for global trade (starting with scrap metal) | 3rd place — Everything AI pitch, NY Tech Week Jun 3 | — |
Five segment trends worth watching
1. The agent security stack is crystallizing
Offroad (identity discovery and OAuth governance), A Security (offensive validation that stress-tests infrastructure continuously), and Tenet Security (runtime protection with pre-execution simulation) are building three distinct layers of the same enterprise anxiety: how do you govern software that acts on its own? Offroad’s public audit finding that one in three marketplace OAuth apps carried structural security concerns — covering more than 1.85 billion installs — provides a data-backed rationale for the stack. Tenet’s “Agentjacking” research, which found thousands of organizations exposed through publicly accessible attack paths that bypass traditional controls because agents were acting within their authorized permissions, extends that rationale into runtime. Expect consolidation within twelve months as large platform vendors seek to acquire all three layers.
2. LLM ops is becoming a commodity — and then isn’t
Concentrate AI’s zero-markup gateway (130+ models, single API, free tier at cost) and Tiger Data’s disposable Postgres environments for agents suggest the routing layer is commoditizing fast. The durable moat will be observability, eval, and cost attribution — not model access. Tenet’s Fortune 1000 deployment, which caught an autonomous agent generating unnecessary token costs over a single weekend, illustrates why: the problem shifts from “which model” to “what did the agent actually do and what did it cost.”
3. Vertical AI is winning on proprietary data, not models
Trunk Tools (10M+ construction documents, deployed on hundreds of active U.S. jobsites, caught a $60,000 electrical change missed outside revision clouds), YieldStack (180+ lender programs, 60+ deal attributes), and Pogo (verified transaction data covering 1 in every 150 U.S. shopping trips) all anchor their claims in industry-specific datasets that foundation-model vendors cannot replicate. Data, not weights, is the differentiator. Dr. Sarah Buchner’s Cortex does not compete with GPT-5.5 on benchmarks; it reads construction drawings in a way no general-purpose model ever will.
4. The IP economy is moving upstream of generation
FanClub AI (rights recognition and licensing infrastructure, backed by Deborah Harpur’s 25-year career negotiating rights for LVMH, Netflix, and 180+ film titles) and Taste Labs (aesthetic evaluation data to reduce “AI slop,” co-led by CRV and Amplify Partners at $18.5M) sit before the prompt is ever typed. As copyright litigation accelerates — 125+ active infringement lawsuits and the $1.5 billion Bartz v. Anthropic settlement — “clean training data” is becoming a procurement requirement, not a marketing claim. Taste Labs’ contrarian bet that standard RLHF degrades aesthetic quality rather than improving it represents a new front in the quality-versus-scale debate that will define the next generation of frontier model training.
5. Agents are becoming storefronts
Shoppable’s Universal Checkout MCP server — protected by four U.S. patents filed in 2011 — and Kernso’s intent-resolution engine preview a web where the checkout flow lives inside the conversation, not on a merchant page. Heather Udo’s early data showing AI-referred visitors converting at multiples of traditional search traffic marks a structural shift in e-commerce economics. Payment rails, not product pages, become the bottleneck, and the neutral infrastructure layer between AI assistants and multi-merchant checkout becomes a strategically valuable position.
What this means for operators and investors
For enterprise buyers, the takeaway is procurement: the security stack (Offroad / A Security / Tenet) and the LLM ops stack (Concentrate AI / Tiger Data) will be evaluated together in 2026 RFPs. Offroad’s OAuth audit alone provides CISO-ready justification for a governance line item that did not exist in 2025 budgets. Bundling conversations early saves procurement cycles later — and having three separate vendors covering identity posture, offensive validation, and runtime protection creates a coherent architecture rather than overlapping point solutions.
For investors, three signals stand out. First, seed rounds in agent security are pricing at Series A multiples — A Security’s $37 million across two rounds, with Lightspeed and Wiz CEO Assaf Rappaport anchoring the follow-on, is a classic pre-consolidation pattern. Second, vertical AI winners are being picked on data exclusivity, not model performance: Trunk Tools’ 10M-document training corpus and Pogo’s $470B transaction panel are moats that a frontier model cannot close by scaling compute. Third, conversational commerce is moving from demo to revenue faster than the 2024–2025 cycle did, because MCP-style standards (Shoppable) remove the integration tax that previously required multi-year platform partnerships.
For founders still in stealth, the window to launch in New York with a credible buyer narrative is open through Q3 2026 — but only if you can point to signed pilots, not just benchmarks. The companies that raised material capital in this wave (Taste Labs at $18.5M, Minerva at $20M, Pogo at $32M total) all entered their public launches with named enterprise customers, production deployments, and in several cases multi-year data assets built during stealth.
Methodology and verification
Profiles follow the PBD Editorial Process. A company is included when it meets at least one of the following between May 19 and June 18, 2026:
- A dated press release or Business Wire distribution with a New York dateline.
- A public product launch or general-availability announcement tied to a verified NYC headquarters.
- A documented appearance at NY Tech Week (June 1–7, 2026) with press coverage.
Verified directory listings carry ten or more cited sources, explicit NAP notes, and disclaimers where street addresses or phone numbers remain unpublished. Report corrections via the individual listing or the PBD contact page.
Related reading
- NYC Enterprise AI in 2026 — February–March 2026 vendor map (prior wave)
- Full places index — all verified PBD profiles
- Editorial process & corrections policy
Frequently asked questions
What counts as “launched in New York” for this list?
We include stealth exits, first public product launches, and major GA announcements with a New York dateline or verified NYC headquarters between May 19 and June 18, 2026. It is a desk-review map, not a legal registry of incorporation dates.
Why do only thirteen companies link to Professional Business Directory listings?
Those thirteen have verified GeoDirectory profiles with sources, NAP notes, and disclaimers. Seven additional companies appear from NY Tech Week and press coverage and are listed separately until dedicated profiles are published.
Is this investment advice?
No. Funding amounts come from public announcements. Verify terms, risks, and fit with qualified advisors before making any investment decision.
Why are many addresses described as “street not listed”?
Early-stage AI startups often publish city-level HQ before a full street address appears on corporate sites. Our listings document what was visible in June 2026 and invite corrections.
How were cybersecurity vs general AI companies categorized?
Identity security, offensive security, and agent runtime protection map to our cybersecurity category (521); infrastructure, research, marketing, and vertical AI map to technology & software (520).
Will this page be updated?
Yes — when new NYC AI launches are verified, when directory profiles gain street addresses, phones, or registry data, or when funding figures are revised by the companies themselves.
How is “disclosed capital” calculated?
We sum only amounts explicitly stated in dated press releases or confirmed by the company. Multi-year totals (e.g., Mecka AI) are flagged as such. Undisclosed rounds are excluded from segment totals.
Last updated (desk review): June 21, 2026 — Cross-checked against PBD place profiles 6742–6754, NY Tech Week coverage (June 1–7, 2026), and selected press items. Funding totals revised to $198.6M disclosed.
Have a correction, a missing launch, or a listing request? Contact the editorial desk.