Fingerprinting survives cookie blocking, and one vendor selling it passed $50M in ARR
TL;DR
Browser fingerprinting identifies a device from hardware and software traits, so clearing cookies or using incognito does not reset it. Google allowed the technique in its ad products from 16 February 2025, and Fingerprint, the main commercial vendor, reports passing $50M in ARR.
On this page
What happened
Browser fingerprinting combines measurements a browser hands over without any permission prompt: GPU model, CPU thread count, screen size, audio processing, time zone, language and installed fonts. Each reading is common on its own. Together they narrow a device to a pool small enough to single it out. Clearing cookies or opening incognito changes none of them.
A post on royfish.dev shows this on the page itself. Per the post, a script takes five measurements (audio rendering, WebGL, canvas drawing, Intl settings and navigator values), hashes them with SHA-256, and returns an animal nickname that stays the same after a refresh, a cookie wipe or a VPN switch. The post also explains the information theory behind it: each trait cuts the pool of candidates by some number of bits, and it cites a ZIP code, birth date and sex as enough to identify over 60% of Americans.
Fingerprints are the new cookies (royfish.dev)The original post. It runs a live fingerprint demo on the page and walks through the entropy math.royfish.dev →The post frames Google's moves as a three-step reversal. The record from other sources supports the outline:
- Google says fingerprinting "subverts user choice and is wrong" (quoted by the ICO).
- Google decides to keep third-party cookies (per LexisNexis).
- Google tells advertisers it will lift its fingerprinting ban (per the ICO, which responded the next day).
- The new policy takes effect. Advertisers using Google's ad products may fingerprint without breaching Google's rules (per the ICO).
Why it matters now
The UK ICO called Google's change irresponsible and said fingerprinting is not a fair way to track people because it reduces their choice and control. Per a DLA Piper summary, the regulator also said fingerprinting needs consent like any other tracking, including when it is used for fraud prevention. Google's policy now allows it. UK law still asks sites to get consent first.
Browser defenses split into two approaches, per the royfish.dev post. Brave adds small amounts of noise (the post calls it farbling) so the fingerprint changes between sessions. Tor and Mullvad Browser make every user look the same. The post names Brave, Tor and Mullvad Browser as the only major browsers with engine-level protection, and says privacy extensions can make a browser more unique because few people run them. That is the author's view, and the post gives no test data for it.
Who sells it
Fingerprint (formerly FingerprintJS Pro) is the main commercial vendor. Its open-source library has more than 28,000 GitHub stars and about 6.3 million monthly downloads, per Sacra, and the paid cloud service sits on top of it. Customers named in company releases include Dropbox, Glovo and Ramp.
| Product | What it sells | Price | Scale |
|---|---|---|---|
| Fingerprint | A visitor ID plus 20+ add-on signals such as bot, VPN and incognito detection | Free: 1,000 calls a month. Pro Plus: $99 a month with 20,000 calls, then $4 per 1,000 | 6,000+ companies, $77M raised (Sacra) |
| cside | A fingerprinting API, pitched as the cheaper option | Free: 1,000 calls. $99 a month with 50,000 calls, then $2 per 1,000 (cside's own comparison) | Not found |
| SEON | Device fingerprint as one input to fraud scoring | Not found | Not found |
| TrustDecision | Device ID and risk labels from 150+ data points, aimed at banks and fintechs | Not found | Not found |
On revenue, a Fingerprint technical lead posted on LinkedIn that the company passed $50 million in ARR, and Sacra estimates $50M in August 2026. Fingerprint's April 2025 release reported 175% revenue growth over two years and 50 new enterprise customers in its fiscal 2025, with no dollar figure. One data aggregator lists $15.4M in revenue and $33M raised. The $33M conflicts with the $77M in the company's own release, so that listing looks stale. No audited confirmation of the $50M exists.
What a small builder could make
These gaps are inferences from pricing and policy pages. No demand has been verified.
Pricing is the first gap. Fingerprint's free plan stops at 1,000 calls a month and its first paid plan costs $99. Sacra lists free-tier abuse at self-serve AI products as a target use case, and small apps with free tiers face the same abuse. A plan at $9 to $29 a month for apps with a few thousand daily visitors would sit below that entry price. cside already charges less per call, so price alone may not hold as an edge.
Consent is the second. The ICO says fingerprinting needs consent in the UK, and Google's change does not alter that. A package that pairs a device ID with a consent banner and a log of each visitor's choice would suit UK and EU sites that want fraud checks without the legal exposure. The searches did not check whether a vendor already bundles the two.
The defense side has independent products too. StealthHound, a Chrome and Edge extension that detects and blocks fingerprinting, launched on Product Hunt in January 2026. Its maker says it reached thousands of users and placed fourth for the day, and a startup directory lists 168 votes. No revenue figure was found. The royfish.dev post argues extensions protect less than engine-level browser defenses, which may cap what this model can charge.
The open-source FingerprintJS library covers the browser side. The royfish.dev post says professional systems keep a visitor's identity when a single trait changes, which suggests the matching logic on the server is where most of the build effort sits. That is an inference, not a measured cost.
What to watch
- The ICO issued draft guidance on storage and access technologies in December 2024, per DLA Piper. Whether it was finalized was not checked.
- Fingerprint's $50M ARR rests on an employee's post and a Sacra estimate. A company announcement or filing would settle it.
- The post's browser rankings come with no test data. The EFF's Cover Your Tracks tool, which the post recommends, shows how unique a given browser is.
Sources
- Fingerprints are the new cookies (royfish.dev): the original post. How fingerprinting works, the five measurements, entropy, browser defenses.
- ICO: our response to Google's policy change: the policy change date and the regulator's position.
- DLA Piper: Google's U-turn on device fingerprinting: the consent requirement and the draft guidance.
- LexisNexis: Google's device fingerprinting U-turn: the July 2024 decision to keep third-party cookies.
- Sacra: Fingerprint: ARR estimate, pricing, GitHub stars, downloads.
- Fingerprint docs: billing and plans: free and paid plan limits.
- Fingerprint April 2025 release: $77M funding, revenue growth, enterprise customers.
- Fingerprint technical lead on LinkedIn: the $50M ARR statement.
- cside: FingerprintJS pricing: cside and Fingerprint price points. A competitor's comparison.
- SEON: device fingerprinting and TrustDecision: device intelligence: what each product does.
- StealthHound on Product Hunt (maker's post) and Innolope listing: launch date, vote count, user and ranking claims.
Continue reading
An off-the-shelf AI model helped XBOW build a Chrome full-chain exploit in six days for $40,000
XBOW says one researcher and AI agents running on OpenAI's publicly available GPT-5.6 Sol built a three-bug Chrome exploit chain in six days, using about $40,000 of compute. Google paid a $250,000 reward, and the bugs are fixed.
Z.ai's GLM-Legal nearly triples its base model's score on Harvey's legal benchmark
Z.ai's GLM-Legal scored 19.2% on Harvey's Legal Agent Benchmark against 6.7% for its GLM-5.3-Flash base, per Z.ai's 12 October release. Vals AI's independent run puts it at 19.17%, seventh of 79 models, at $0.15 per test.
Gemini 4 Argon vs Muse Wonder: same $4/$20 list price, different access and strengths
Google's Gemini 4 Argon and Meta's Muse Wonder both list at $4 per million input tokens and $20 per million output tokens. Argon costs half that through November, and Meta's stronger Wonder Full is not sold through its API.