The landscape · Cybersecurity
Cybersecurity: what AI is really doing
Brief me in 2 minutes
AI in Cybersecurity.
AI in cybersecurity right now means narrow, supervised automation: agents that hunt for exploitable bugs, triage SOC alerts, and vet the plug-ins other AI agents install, with a human still signing off before anything gets patched or shipped. Gartner says companies will spend $2.83 billion this year securing AI tools and agents, 83% more than 2025, because the same generative models that write phishing emails and clone voices are also the ones finding software vulnerabilities before a person does. The pitch behind giant rounds like Kevin Mandia's $189.9 million bet, autonomous AI armies fighting autonomous AI armies, is still mostly a pitch: what is actually shipping right now is agents that make one security team faster, not agents that replace it.
The game: we show you 6 things happening. You call each one real now, early or mostly hype. Then the receipts.
Shift 1 of 6
Autonomous AI agents take over penetration testing and vulnerability hunting
Real now
Horizon3 has run 310,000 production tests, crossed $100 million in annual recurring revenue, and just tripled its valuation to $2 billion, that is paying customers renewing, not a slide deck.
How it works. Horizon3's NodeZero platform runs AI agents that continuously probe a company's own network for exploitable holes the way an attacker would, then hands operators a ranked list of real weaknesses instead of the once-a-year sample a human pentest firm delivers. It tests live systems without shutting them down, repeating the scan on an ongoing basis rather than annually.
Overstated: Horizon3 calls its product "AI Hackers" running fully autonomous penetration testing, but a human still authorizes every test and reviews the ranked list of weaknesses before anyone patches anything.
Shift 2 of 6
Security operations centers hand first-line alert triage to AI agents
Early
Exaforce has real paying customers like Replit and Guardant Health, but the startup only shipped its product a year ago and is still counting customers in the dozens, not the thousands.
How it works. Exaforce runs four specialized AI agents, its "Exabots," that pull raw data out of a company's security tools, correlate it, and decide which alerts are real attacks versus noise, the job that used to fall to an overworked human analyst reading hundreds of alerts a shift. Analysts can then query the system in plain English to chase a hunch instead of writing a formal search.
Overstated: Exaforce brought its product to market only a year ago and counts customers in the dozens so far, so the 90% workload-reduction claim is the vendor's own number from early deployments, not an independently audited industry figure.
Shift 3 of 6
Criminals weaponize AI for phishing and voice-clone fraud, so defenders fight AI with AI
Real now
These are not lab demos: AegisAI and Ocean are already scanning production email for paying customers including Kayak and LangChain, and Savi's consumer app is live on the app stores.
How it works. Criminals now use AI agents to scrape a target's coworkers, active projects, and travel plans, then auto-generate a bespoke spear-phishing email or a cloned voice call in minutes instead of the days it used to take a skilled human attacker. Startups like AegisAI and Ocean fight back by running AI agents that read every message the way a suspicious analyst would, instead of matching it against a fixed rulebook.
Overstated: AegisAI and Ocean both say their agents catch what rule-based filters miss, but neither claims to stop AI attacks entirely, and AegisAI's own figures say AI-crafted attacks still slip past existing defenses more than half the time.
Shift 4 of 6
Deepfake video and voice fraud hits the wire-transfer desk, so defense shifts to real-time identity checks
Early
The Zoom integration just shipped and only protects people who opted into World's ID system, while the fraud technique it defends against has already been costing real companies real money for two years.
How it works. Fraud crews now stage entire fake video calls, cloning a CFO's face and voice in real time to authorize wire transfers, which is why platforms like Zoom are adding live identity verification instead of only scanning recorded frames for manipulation after the fact. Sam Altman's World checks a registered face scan, a live face scan, and the live video feed together before it marks anyone "Verified Human."
Overstated: World itself says the frame-by-frame deepfake detection most vendors still sell is becoming unreliable as video models improve, and its own face-scan alternative only protects people who already enrolled in World's ID system.
Shift 5 of 6
Enterprises rush to secure the AI agents and plug-ins they just deployed, not just the model
Early
The spending and the funding rounds are real and large, but by the vendors' own admission there are still few confirmed cases of an AI agent being exploited in production.
How it works. As companies give AI agents access to internal systems, a new supply chain of skills, plug-ins, and MCP servers has sprung up around them, largely unsigned and unvetted, the way software drivers went unsigned before code-signing existed. Startups like AIR and HiddenLayer now sit between an agent and everything it touches, filtering malicious tools, prompt injection, and poisoned content before the agent acts on them.
Overstated: HiddenLayer's own CEO admits there still aren't many public headlines about AI agents actually being exploited at scale, this is money moving ahead of documented damage, not behind it.
Shift 6 of 6
VCs write record-size early checks for AI-native security startups
Mostly hype
Both rounds are real money in the bank, but neither company has published a customer count, a revenue number, or independent proof the product works at scale, that is a bet on the founders, not on receipts.
How it works. Investors are funding AI security startups before they have proven revenue at a scale that would normally justify it: Kevin Mandia's new startup raised what it calls a record combined seed and Series A for a security company without disclosing a valuation, and Glow raised a Series A that made it a unicorn while still declining to disclose customer numbers. The bet is that whoever builds the trusted layer between AI agents and the enterprise first wins the category.
Overstated: Armadin has not disclosed a valuation or public product results, and Glow only reports "paying customers" without naming a single one, big rounds are not the same thing as a proven market.
The scary part
What happens to jobs in Cybersecurity
What AI is already taking:
- Reading every inbound email for AI-crafted spear-phishing and impersonation before it reaches an inbox TechCrunch ↗
- Running the network penetration test that used to take a human security firm a scheduled week each year TechCrunch ↗
- Triaging which of hundreds of daily SOC alerts is a real, high-priority attack TechCrunch ↗
- Drafting security policies and collecting the evidence auditors need for a SOC 2 report TechCrunch ↗
What stays human:
A person still signs off before an AI-flagged vulnerability gets patched, an AI-drafted policy ships, or an incident gets reported to a regulator, because every vendor above builds it that way on purpose. What stays human is judgment under real ambiguity, deciding whether a genuinely new attack pattern is real or noise, which is exactly the case a model trained on past attacks handles worst.
Get better · 15 minutes
The ability to read an AI security agent's reasoning and catch the case where it is confidently wrong, not the ability to write detection rules by hand.
Try this this week:
Take a phishing message you've already confirmed as malicious this week and run it through Savi's free Scamwise checker at scamwise.com alongside one message you know is clean, then compare its verdict to your own call. Do this for three messages, about 15 minutes, and you'll see exactly where an AI scam-detector's confidence does and doesn't match reality.
Your move
Model-and-agent security (AIR, HiddenLayer, Zenity, Noma) is already crowded with well-funded players chasing the same MCP-and-plug-in vetting problem; the emptier space is proving an AI defender's own decision was correct to a regulator or insurer after the fact, nobody yet sells that audit trail.
Ask for a named paying customer and one real attack the tool actually caught, not a headline valuation, Armadin and Glow both raised nine-figure rounds without disclosing a customer name or a revenue number, and that gap is exactly what a term sheet should be pricing.
This quarter, feed your own SOC's alert backlog into one AI triage tool on a trial for two weeks and compare its flagged list against what your team actually escalated, that gap tells you whether the tool is ready to run unsupervised or only ready to summarize.
Learn to audit an AI security agent's decision trail rather than only writing detection rules by hand, the analysts becoming harder to replace are the ones who can explain why an agent flagged or missed something, starting with a week of your own SOC's AI-triaged alerts reviewed end to end.
What's actually changing
- 1
Autonomous AI agents take over penetration testing and vulnerability hunting
How it works. Horizon3's NodeZero platform runs AI agents that continuously probe a company's own network for exploitable holes the way an attacker would, then hands operators a ranked list of real weaknesses instead of the once-a-year sample a human pentest firm delivers. It tests live systems without shutting them down, repeating the scan on an ongoing basis rather than annually.
Why it matters. Companies that used to test their defenses once a year on a slice of their systems can now do it continuously across the whole network, which is often the difference between finding a hole before or after someone else does.
What's overstatedHorizon3 calls its product "AI Hackers" running fully autonomous penetration testing, but a human still authorizes every test and reviews the ranked list of weaknesses before anyone patches anything.
Horizon3.ai
3 receipts
- Horizon3 raised a $250 million Series E at a $2 billion valuation, more than tripling its valuation in 14 months.
Cybersecurity startup Horizon3 has raised a $250 million Series E at a $2 billion valuation, more than tripling its valuation in 14 months.
TechCrunch, 3 Aug 2026 ↗ - Horizon3 has run 310,000 production security tests with zero disruptions under what it calls fully autonomous "AI Hackers" penetration testing.
In a statement, Horizon3 has run 310,000 production security tests with zero disruptions, establishing what it calls "AI Hackers," fully autonomous penetration testing.
TechCrunch, 3 Aug 2026 ↗ - Horizon3 approached $100 million in annual recurring revenue last year with 120% year-over-year growth.
The company approached $100 million in annual recurring revenue last year with 120% year-over-year growth, and expects to accelerate growth this year, Hartley said.
TechCrunch, 3 Aug 2026 ↗
- Horizon3 raised a $250 million Series E at a $2 billion valuation, more than tripling its valuation in 14 months.
- 2
Security operations centers hand first-line alert triage to AI agents
How it works. Exaforce runs four specialized AI agents, its "Exabots," that pull raw data out of a company's security tools, correlate it, and decide which alerts are real attacks versus noise, the job that used to fall to an overworked human analyst reading hundreds of alerts a shift. Analysts can then query the system in plain English to chase a hunch instead of writing a formal search.
Why it matters. Security teams drowning in false-positive alerts get to spend their time on the handful that matter, and Exaforce says its platform strips out as much as 90% of that manual grind.
What's overstatedExaforce brought its product to market only a year ago and counts customers in the dozens so far, so the 90% workload-reduction claim is the vendor's own number from early deployments, not an independently audited industry figure.
Exaforce7AIDropzone AIProphet Security
3 receipts
- Exaforce, a three-year-old AI startup that detects and stops cyberattacks in real time, raised a $125 million Series B.
Exaforce, an AI startup that detects and thwarts attacks in real time, secure a $125 million Series B.
TechCrunch, 12 May 2026 ↗ - The round valued Exaforce at $725 million, with backing from HarbourVest, Peak XV, Mayfield, Khosla Ventures, and Seligman Ventures.
The round valued the three-year-old startup at $725 million and saw participation from HarbourVest, Peak XV, Mayfield, Khosla Ventures, and Seligman Ventures.
TechCrunch, 12 May 2026 ↗ - Exaforce claims its AI platform can reduce manual, time-consuming SOC tasks by as much as 90%.
Exaforce claims its AI platform can reduce manual, time-consuming tasks by as much as 90%.
TechCrunch, 12 May 2026 ↗
- Exaforce, a three-year-old AI startup that detects and stops cyberattacks in real time, raised a $125 million Series B.
- 3
Criminals weaponize AI for phishing and voice-clone fraud, so defenders fight AI with AI
How it works. Criminals now use AI agents to scrape a target's coworkers, active projects, and travel plans, then auto-generate a bespoke spear-phishing email or a cloned voice call in minutes instead of the days it used to take a skilled human attacker. Startups like AegisAI and Ocean fight back by running AI agents that read every message the way a suspicious analyst would, instead of matching it against a fixed rulebook.
Why it matters. Attacks that used to require a well-resourced attacker's patience are now cheap enough to run against ordinary employees, and the FTC says imposter-scam losses reported by consumers roughly tripled in five years.
What's overstatedAegisAI and Ocean both say their agents catch what rule-based filters miss, but neither claims to stop AI attacks entirely, and AegisAI's own figures say AI-crafted attacks still slip past existing defenses more than half the time.
AegisAIOceanSavi Security
3 receipts
- AegisAI raised a $36 million Series A to build AI agents that catch AI-crafted spear-phishing emails.
helped AegisAI raise a $36 million Series A led by Battery Ventures, with participation from existing backers Accel and Foundation Capital.
TechCrunch, 23 July 2026 ↗ - AegisAI's co-founder says AI-crafted attacks now bypass existing email defenses more than half the time, roughly twice as often as before.
AI-powered attacks bypass existing controls more than half the time now, which means they're almost twice as effective as they used to be.
TechCrunch, 23 July 2026 ↗ - Ocean, an agentic email security platform, emerged from stealth with $28 million in total funding to fight AI-powered phishing.
His startup Ocean, an agentic email security platform built to fight AI-powered attacks, just emerged from stealth mode with $28 million in total funding.
TechCrunch, 19 May 2026 ↗
- AegisAI raised a $36 million Series A to build AI agents that catch AI-crafted spear-phishing emails.
- 4
Deepfake video and voice fraud hits the wire-transfer desk, so defense shifts to real-time identity checks
How it works. Fraud crews now stage entire fake video calls, cloning a CFO's face and voice in real time to authorize wire transfers, which is why platforms like Zoom are adding live identity verification instead of only scanning recorded frames for manipulation after the fact. Sam Altman's World checks a registered face scan, a live face scan, and the live video feed together before it marks anyone "Verified Human."
Why it matters. The fraud already moves real money: losses topped $200 million in a single quarter by one estimate and now average $500,000 per corporate incident, so any company still authorizing a wire transfer on a video call alone is trusting something that can now be faked live.
What's overstatedWorld itself says the frame-by-frame deepfake detection most vendors still sell is becoming unreliable as video models improve, and its own face-scan alternative only protects people who already enrolled in World's ID system.
WorldZoom
3 receipts
- Financial losses from deepfake-enabled fraud exceeded $200 million in a single quarter, and the average loss per corporate incident now tops $500,000.
financial losses from deepfake-enabled fraud exceeded $200 million in just the first quarter of last year, according to one estimate, and the average loss per corporate incident now tops $500,000, according to security industry reports.
TechCrunch, 17 Apr 2026 ↗ - Zoom and World verify a meeting participant as human only when a registered face scan, a live face scan, and the live video feed all match.
It only verifies someone when all three things match, at which point a "Verified Human" badge appears on that participant's title.
TechCrunch, 17 Apr 2026 ↗ - World says older deepfake detection, which analyzes video frames for manipulation, is becoming unreliable as video-generation models improve.
because video models are getting better, those frame-by-frame detection methods are increasingly unreliable
TechCrunch, 17 Apr 2026 ↗
- Financial losses from deepfake-enabled fraud exceeded $200 million in a single quarter, and the average loss per corporate incident now tops $500,000.
- 5
Enterprises rush to secure the AI agents and plug-ins they just deployed, not just the model
How it works. As companies give AI agents access to internal systems, a new supply chain of skills, plug-ins, and MCP servers has sprung up around them, largely unsigned and unvetted, the way software drivers went unsigned before code-signing existed. Startups like AIR and HiddenLayer now sit between an agent and everything it touches, filtering malicious tools, prompt injection, and poisoned content before the agent acts on them.
Why it matters. Gartner expects companies to spend $2.83 billion this year securing AI tools and agents, 83% more than 2025, because the blast radius of one poisoned plug-in spreading across an entire agent fleet is enormous even while confirmed exploits stay rare.
What's overstatedHiddenLayer's own CEO admits there still aren't many public headlines about AI agents actually being exploited at scale, this is money moving ahead of documented damage, not behind it.
AIRHiddenLayerZenityNoma SecurityComp AI
3 receipts
- Gartner estimates companies will spend $2.83 billion this year securing AI tools, 83% more than 2025, and expects that to reach nearly $4.78 billion next year.
Gartner estimates companies will spend $2.83 billion this year on products meant to secure AI tools, 83% more than 2025, and expects spending to reach nearly $4.78 billion next year.
TechCrunch, 2 Sept 2026 ↗ - AIR came out of stealth with $50 million raised across two seed rounds to vet the skills, plug-ins, and MCP servers AI agents use.
AI security startup AIR believes companies will need a way to monitor that supply chain, and it's now coming out of stealth with $50 million raised across two seed rounds
TechCrunch, 1 Sept 2026 ↗ - Zenity raised a $125 million Series C and Noma raised a $100 million Series B, both competing with AIR and HiddenLayer in AI-agent security.
Zenity raised a $125 million Series C in August, while Noma raised a $100 million Series B last year.
TechCrunch, 1 Sept 2026 ↗
- Gartner estimates companies will spend $2.83 billion this year securing AI tools, 83% more than 2025, and expects that to reach nearly $4.78 billion next year.
- 6
VCs write record-size early checks for AI-native security startups
How it works. Investors are funding AI security startups before they have proven revenue at a scale that would normally justify it: Kevin Mandia's new startup raised what it calls a record combined seed and Series A for a security company without disclosing a valuation, and Glow raised a Series A that made it a unicorn while still declining to disclose customer numbers. The bet is that whoever builds the trusted layer between AI agents and the enterprise first wins the category.
Why it matters. When investors write nine-figure checks into pre-revenue-disclosure security startups, it signals a bet that AI-native security will be a winner-take-most category worth defending early, and also that a good chunk of that money will fund products that never find a real buyer.
What's overstatedArmadin has not disclosed a valuation or public product results, and Glow only reports "paying customers" without naming a single one, big rounds are not the same thing as a proven market.
ArmadinGlow
3 receipts
- Kevin Mandia's new startup Armadin raised $189.9 million in combined seed and Series A funding led by Accel.
The new outfit, called Armadin, has raised $189.9 million in combined seed and Series A funding led by Accel
TechCrunch, 10 Mar 2026 ↗ - Mandia warns that AI on offense is a technology that can think, learn, and adapt to complete attacks that used to take days in minutes.
When you have AI on offense, what you are going to get is a technology that can think, can learn, can adapt
TechCrunch, 10 Mar 2026 ↗ - Glow raised $180 million in an all-equity Series A that valued the endpoint security startup at $1.2 billion.
The Palo Alto-headquartered startup on Wednesday said it raised $180 million in an all-equity Series A funding round that valued it at $1.2 billion
TechCrunch, 22 July 2026 ↗
- Kevin Mandia's new startup Armadin raised $189.9 million in combined seed and Series A funding led by Accel.
Who's doing it
13 companies to know, 13 with a round we can source. Newest first where dated.
Search these companies in the startup database ↗ · Explore all research tools
Comp AIagentic security and compliance automation
AI agents draft security policies, collect audit evidence, and continuously check whether a company still meets compliance controls after it deploys new AI agents.
Latest · 17 Sept 2026Comp AI has raised $37.5 million in total funding to date, most recently a $34 million Series A. TechCrunch ↗
$34M Series A, 17 Sept 2026 source ↗
HiddenLayerAI model, agent, and supply-chain security
Protects AI models, agents, and workflows from adversarial attacks, prompt injection, and malicious code hidden inside AI file formats.
Latest · 2 Sept 2026One of HiddenLayer's customers is apparently a leading frontier model provider with more than 700 million weekly users, which its CEO says sounds like OpenAI or Anthropic. TechCrunch ↗
$100M Series B, 2 Sept 2026 source ↗
AIRAI agent supply-chain security (skills, plug-ins, MCP servers)
Sits between an AI agent and the tools it uses, continuously vetting skills, plug-ins, and MCP servers and blocking ones that fail its security checks.
Latest · 1 Sept 2026AIR's platform currently filters out about 27% of the add-ons and skills it finds online, weeding out ones that turn risky after approval. TechCrunch ↗
$50M two seed rounds, 1 Sept 2026 source ↗
ZenityAI agent security and governance
Security and governance tools for AI agents, coding agents, and MCP servers running across a company's SaaS and cloud environment.
Latest · 1 Sept 2026Zenity raised a $125 million Series C in August 2026, one of the largest rounds yet in AI agent security. TechCrunch ↗
$125M Series C, 1 Sept 2026 source ↗
Noma SecurityAI application and agent security
A unified platform for discovering and securing the LLMs, RAG pipelines, and AI agents a company's employees actually use.
Latest · 1 Sept 2026Noma offers discovery, access controls, and runtime monitoring for agents, MCP servers, and skills, and raised a $100 million Series B. TechCrunch ↗
$100M Series B, 1 Sept 2026 source ↗
Horizon3.aiautonomous penetration testing / vulnerability discovery
Runs AI agents (NodeZero) that continuously and safely probe a company's own network for exploitable vulnerabilities the way an attacker would.
Latest · 3 Aug 2026Horizon3 has run 310,000 production security tests with zero disruptions under its fully autonomous "AI Hackers" platform. TechCrunch ↗
$250M Series E, 3 Aug 2026 source ↗
AegisAIAI-native email / anti-phishing security
AI agents that read incoming email like a suspicious analyst would, to catch AI-crafted spear-phishing that rule-based filters miss.
Latest · 23 July 2026AegisAI says its tech has been adopted by dozens of customers, including crypto payments company Mesh, AI startup LangChain, and privacy compliance platform Lokker, within a year of launch. TechCrunch ↗
$36M Series A, 23 July 2026 source ↗
GlowAI-era endpoint security
An endpoint security platform that uses AI agents to map the software, AI agents, and developer tools running on employee devices, and blocks risky ones before they install.
Latest · 22 July 2026Glow's platform has already prevented malicious npm packages from being installed in customer environments and flagged devices with missing or degraded endpoint protection. TechCrunch ↗
$180M Series A, 22 July 2026 source ↗
Savi Securityconsumer AI-scam and deepfake defense
A consumer app and live-call agent that listens for the behavioral tells of an AI-generated scam call, voice clone, or deepfake kidnapping hoax in real time.
Latest · 7 July 2026The FTC says people reporting online crimes collectively lost $3.5 billion to imposter scams in 2025, triple the amount lost in 2020. TechCrunch ↗
$7M seed, 7 July 2026 source ↗
Oceanagentic email security
An agentic email security platform built to catch AI-generated phishing that mimics a person's real writing and organizational context.
Latest · 19 May 2026Ocean is already reviewing billions of emails each month for customers including Kayak, Kingston Technology, and Headspace. TechCrunch ↗
$28M seed (total, at stealth exit), 19 May 2026 source ↗
ExaforceAI-native SOC / detection and response
Runs four AI agents, its "Exabots," that triage security alerts and let analysts query for threats in plain English.
Latest · 12 May 2026Exaforce claims its AI platform can reduce manual, time-consuming SOC tasks by as much as 90%, with customers including Replit and Guardant Health. TechCrunch ↗
$125M Series B, 12 May 2026 source ↗
Armadinautonomous AI agent security
Kevin Mandia's new startup, building autonomous AI agents meant to detect and respond to cyberattacks without a human in the middle.
Latest · 10 Mar 2026Armadin raised $189.9 million in combined seed and Series A funding, which it says is a record for a security startup at that early a stage. TechCrunch ↗
$189.9M Seed + Series A, 10 Mar 2026 source ↗
DepthfirstAI-native code and workflow security scanning
An AI-native platform, General Security Intelligence, that scans codebases and workflows for credential exposure and threats from open-source dependencies.
Latest · 14 Jan 2026Depthfirst points to Anthropic's claim that it thwarted an AI orchestrated cyber espionage campaign as evidence the threat it defends against is already real. TechCrunch ↗
$40M Series A, 14 Jan 2026 source ↗
Your move
Our read, built on the receipts above.
Model-and-agent security (AIR, HiddenLayer, Zenity, Noma) is already crowded with well-funded players chasing the same MCP-and-plug-in vetting problem; the emptier space is proving an AI defender's own decision was correct to a regulator or insurer after the fact, nobody yet sells that audit trail.
Ask for a named paying customer and one real attack the tool actually caught, not a headline valuation, Armadin and Glow both raised nine-figure rounds without disclosing a customer name or a revenue number, and that gap is exactly what a term sheet should be pricing.
This quarter, feed your own SOC's alert backlog into one AI triage tool on a trial for two weeks and compare its flagged list against what your team actually escalated, that gap tells you whether the tool is ready to run unsupervised or only ready to summarize.
Learn to audit an AI security agent's decision trail rather than only writing detection rules by hand, the analysts becoming harder to replace are the ones who can explain why an agent flagged or missed something, starting with a week of your own SOC's AI-triaged alerts reviewed end to end.
Other industries
- FinanceBanks move agentic AI from chatbots to running back-office and dealmaking work
- Software engineeringSpaceX buys the leading AI coding startup outright, for $60 billion
- HealthcareAmbient scribes go from pilot project to clinical infrastructure
- LegalLegal AI vendors go from research assistant to agent that produces the deliverable
- Real estateLeasing and resident calls move to voice AI agents that never sleep
- Sales and marketingCustomer support agents now get paid per resolved ticket, not per seat
- Jobs, HR and recruitingAI runs the first-round interview, not just the resume filter
- InsuranceClaims move from adjuster-first to AI-first, humans supervise the exceptions
- Consulting and accountingAI-written client reports become a recurring, public Big Four embarrassment
- Retail and e-commerceShopping agents get a checkout button, and Amazon spends 2026 deciding who's allowed to press it
- EducationNYC and LA ban AI for students, not for teachers
- Media and creativeAI video splits into model labs that train and aggregators that resell
- Manufacturing and roboticsHumanoid robots move from demo livestreams to paid warehouse shifts
- Transport and robotaxisRobotaxis go from demo to metered fare across a dozen-plus US cities
- Energy and data-centre powerGrid rules get rewritten mid-buildout, state by state
- GovernmentGSA turns AI adoption into a government-wide subscription, then flips 'free' to metered
- Agriculture and foodAutonomous tractor fleets learn to run the tools behind them, not just the wheel