Home/Reports/BioLLM (Ankythera AI)
Last reviewed June 22, 2026FILE №BIOLLM
AI CRYPTO PRODUCTAI · CRYPTO · BIOCOMPUTING · DEEP CHECK

Is BioLLM (Ankythera AI) safe to buy from?

The chat works. But you shouldn't upload your DNA to a Florida LLC with no named founders, and you shouldn't expect a Pump.fun token to preserve value. Verdict: Avoid · 40/100.

CrazyCheck Agent v1.0Published June 22, 20267 green9 red flagsConfidence medium
SCORE
0/100
Avoid
● First audit · score 40/100
The verdict

BioLLM presents itself as "the world's first living language model," implying biological neural intelligence. In reality it is a fine-tuned Qwen3.5-4B (Alibaba Cloud open-source LLM) trained on 24,000 stimulation trials from Cortical Labs CL1 MEA experiments; production inference runs a 10-million Izhikevich neuron simulation — a mathematical model, not live neurons. The product works as a chat interface. The operator, Stormy IT Services LLC (DBA Ankythera AI, Florida LLC), is completely anonymous: no founders named, no physical address, contact only via @BioLLMbot. The $BioLLM token was launched on Pump.fun — a Solana memecoin launchpad where 98.6% of tokens show fraud signals — and is down 66% from its March 2026 ATH. OpenCure, the platform's genomic data tool, solicits raw DNA file uploads from users with no named physician, geneticist, or IRB oversight. The "BioLLM Spark" EEG headset listed on the homepage shows no shipping date, no regulatory clearance, and no evidence of existing hardware. The transparency page does honestly disclose that production uses a neuron simulation rather than live cells — a significant credit — but it cannot offset anonymous team risk, memecoin tokenomics, and sensitive personal data collection at scale.

  • Domain registered October 28, 2024; ~1.6 years operating.
  • Florida LLC (Stormy IT Services LLC, DBA Ankythera AI) is disclosed in the Terms of Service with Florida governing law. A legal entity exists — this is not a pure ghost operation with zero corporate registration.
  • The $BioLLM token was launched on Pump.fun, a Solana memecoin launchpad with no vetting, no compliance review, and no project requirements. Per Solidus Labs research (2024), 98.6% of tokens launched on Pump.fun exhibit at least one fraud signal (wash trading, rug pull, coordinated dev wallet dump). The contract address suffix "pump" and the tokenomics page's own disclosure ("Launch Platform: pump.fun") confirm this origin.
  • $BIOLLM is down approximately 66% from its all-time high of $0.002426 reached on March 8, 2026. At audit date the token trades at ~$0.000811 with a market cap of ~$811K and minimal daily volume. Token price action is consistent with post-ATH Pump.fun lifecycle decay.
  • OpenCure solicits raw genomic data uploads (.vcf, .23andme) from an anonymous operator with no named clinical professionals, no IRB approval, no CLIA certification, and no FDA regulatory framework cited. Genomic data is uniquely sensitive and irreversible — once shared, it cannot be recalled. Retaining it for "model improvement" with no oversight is a severe data-risk red flag.

9 red flags

  • The brand headline "The World's First Living Language Model" is materially misleading. The model does not run on living neurons. Production uses a mathematical simulation (Izhikevich model) trained on data from biological experiments. The word "living" in the product name creates a false impression that is only corrected deep in a separate transparency page that most users will not read.
  • The operator is completely anonymous. No founder names, no executive team, no LinkedIn profiles, no company address, no investor deck, no public filings beyond a Florida LLC registered under "Stormy IT Services LLC." The sole public contact for a platform collecting genomic data and selling crypto tokens is an automated bot: @BioLLMbot on X and Telegram.
  • WHOIS privacy is enabled (Namecheap). The operator has taken active steps to prevent public identification of the registrant behind biollm.com, consistent with the platform's broader anonymity posture.
  • The $BioLLM token was launched on Pump.fun, a Solana memecoin launchpad with no vetting, no compliance review, and no project requirements. Per Solidus Labs research (2024), 98.6% of tokens launched on Pump.fun exhibit at least one fraud signal (wash trading, rug pull, coordinated dev wallet dump). The contract address suffix "pump" and the tokenomics page's own disclosure ("Launch Platform: pump.fun") confirm this origin.
  • $BIOLLM is down approximately 66% from its all-time high of $0.002426 reached on March 8, 2026. At audit date the token trades at ~$0.000811 with a market cap of ~$811K and minimal daily volume. Token price action is consistent with post-ATH Pump.fun lifecycle decay.
  • No independent press coverage found at audit date — no CoinDesk, Decrypt, TechCrunch, VentureBeat, or mainstream AI/crypto media coverage. No Crunchbase profile. No Product Hunt listing. No academic mention of Ankythera AI or BioLLM in computational neuroscience literature. All public discourse routes through an automated bot account.
  • OpenCure solicits raw genomic data uploads (.vcf, .23andme) from an anonymous operator with no named clinical professionals, no IRB approval, no CLIA certification, and no FDA regulatory framework cited. Genomic data is uniquely sensitive and irreversible — once shared, it cannot be recalled. Retaining it for "model improvement" with no oversight is a severe data-risk red flag.
  • The "BioLLM Spark" EEG headset appears to be vaporware. No shipping timeline, no prototype evidence, no manufacturing partner, no FDA 510(k) pre-submission, and no independent coverage found at audit date. It is listed on the homepage alongside live products without any qualifier.
  • Payments are exclusively in crypto (SOL, USDC, $BioLLM) with no fiat option and no published refund policy. This removes standard consumer protections available under credit card chargebacks and increases financial risk for paid subscribers, particularly those using the proprietary $BioLLM token (down 66% from ATH) as payment.

7 green signals

  • Florida LLC (Stormy IT Services LLC, DBA Ankythera AI) is disclosed in the Terms of Service with Florida governing law. A legal entity exists — this is not a pure ghost operation with zero corporate registration.
  • The transparency page accurately discloses the real technology stack: Qwen3.5-4B base model, fine-tuned on Cortical Labs CL1 MEA data, with Izhikevich neuron simulation in production. This is honest technical documentation that contradicts the misleading "living language model" headline.
  • Domain biollm.com was registered 2024-10-28 (~20 months at audit date), predating the March 2026 token ATH. The site is fully functional with multiple product sections (Chat, OpenCure, Tokenomics, Pricing, Transparency, Developer API, Staking). This is not a last-minute launch domain.
  • The AI chat product functions: the site is live, the free tier (5 messages/day/IP) requires no wallet or email signup, and Qwen3.5-4B is a capable open-source base model (Alibaba Cloud, Apache 2.0). The product delivers real value as an AI assistant independent of the neuroscience marketing.
  • The transparency page is a genuine disclosure: it explicitly states that production does not use live neurons, names the base model (Qwen3.5-4B), identifies the training data source (Cortical Labs CL1 MEA), and describes the Izhikevich simulation framework. This level of technical transparency is rare for a crypto-AI project.
  • Cortical Labs and the CL1 platform are real, commercially available, and credibly cited. The Izhikevich neuron model is a real computational neuroscience formulation. The technical claims, while unverified by the named third party, are scientifically coherent — not pseudoscience.
  • Mint authority for $BioLLM is revoked and buy/sell tax is 0%. These tokenomics parameters are verifiable on-chain and reduce (though do not eliminate) certain rug pull vectors. The operator cannot mint new tokens to dilute holders.
Footprint50/100
Consensus25/100
Results50/100
The transparency page is honest. The homepage headline is not — and there's no named operator to hold to either one.CrazyCheck.ai editorial, June 2026
01

What BioLLM actually is

Visit biollm.com and the first thing you read is "The World's First Living Language Model — BioLLM bridges biological neural intelligence with advanced language modeling, creating an AI that thinks more like you." It's a striking claim. Biological neural intelligence. A model that thinks. The implied promise is an AI with something qualitatively different going on under the hood — a step beyond the transformer architectures that power every major language model today.

Here is what is actually running: Qwen3.5-4B, an open-source model released by Alibaba Cloud under the Apache 2.0 license, fine-tuned on 24,000 stimulation trials captured from Cortical Labs CL1 multi-electrode array (MEA) experiments. In production, when you send a message, it is processed not by living neurons but by a 10-million-node Izhikevich spiking neuron simulation — a mathematical model developed by computational neuroscientist Eugene Izhikevich in 2003 that replicates the firing dynamics of biological neurons as differential equations. It is real technology. It is not biology.

To their credit, BioLLM's transparency page states this clearly: "BioLLM does NOT run on live biological neurons in production. Inference uses a 10 million Izhikevich neuron simulation." That sentence is accurate and important. The problem is that the homepage headline — the phrase most users will read and most journalists will quote — says the opposite.

  • Branding: "World's First Living Language Model"
  • Reality: Qwen3.5-4B + a 10M-node Izhikevich simulation — not live neurons
02

The operator: a Florida LLC with no face

The legal entity behind BioLLM is Stormy IT Services LLC, a Florida limited liability company doing business as Ankythera AI. This is disclosed in the Terms of Service. That is all that is disclosed. There are no named founders, no executive team, no LinkedIn profiles, no investor deck, no physical address, no email address on the terms page, and no public regulatory filings that would identify the humans operating this platform.

The sole public contact is @BioLLMbot — an automated bot on X (Twitter) and Telegram. WHOIS privacy is enabled on biollm.com (registered 2024-10-28 via Namecheap), so even the domain registration record reveals nothing. The domain is 20 months old, predating the March 2026 token peak, which rules out a pure opportunistic rush-launch. But age does not substitute for identity.

This anonymity matters for several reasons. The platform offers a genomic data tool (OpenCure) that accepts raw DNA files. It operates a cryptocurrency token ($BioLLM) sold to retail holders. It stores user data for "model improvement." Under each of these activities, the question "who is responsible?" has no answer that can be verified by any user of the platform.

  • Entity: Stormy IT Services LLC (DBA Ankythera AI), Florida
  • No named founders, no address, no email — contact is @BioLLMbot only
03

The Pump.fun token

The $BioLLM token is a Solana SPL token launched on Pump.fun (contract address: 9oxGicd5KXNcthYM6yD23X9zcznQVnHjrAEN38fFpump — the "pump" suffix is not a coincidence, it is the standard Pump.fun contract identifier pattern). Supply is 1 billion tokens. Mint authority is revoked. Tax is 0%.

Pump.fun is a permissionless Solana launchpad: anyone can create a token in seconds with no vetting, no compliance review, and no project requirements. Per Solidus Labs research published in 2024, 98.6% of tokens created on Pump.fun show at least one fraud signal — wash trading, coordinated developer wallet dumps, or structured rug pulls. Launching on Pump.fun does not prove fraud. It is, however, a launch context that provides essentially zero signal about project legitimacy and very strong prior odds of value decay.

$BIOLLM reached its all-time high of $0.002426 on March 8, 2026. At the time of this audit it trades at approximately $0.000811 — a 66% decline. Market cap sits around $811,000 with minimal volume. The mint revocation and zero-tax structure are positive parameters that reduce specific rug-pull vectors, but they do not change the trajectory or the launchpad context.

Users can pay for premium BioLLM subscriptions in $BioLLM tokens. This creates a functional loop in which the token's utility is circular: the platform sells tokens, tokens are used to pay the platform. There is no refund policy published for any subscription tier, including those denominated in a volatile asset.

  • $BioLLM launched on Pump.fun (98.6% of its tokens show fraud signals)
  • Down ~66% from its March 2026 ATH; circular token utility, no refunds
04

OpenCure: your DNA, their model

OpenCure is BioLLM's genomic health tool. The pitch: upload your raw genomic data file (.vcf format, 23andMe export, or similar), and receive "personalized health insights powered by BioLLM." The privacy policy states data is "not sold to third parties" but is retained for "model improvement."

Genomic data is among the most sensitive categories of personal information that exists. It is permanent, it is uniquely identifying, it reveals predispositions to disease, it can affect family members who never consented to sharing, and it cannot be changed if mishandled. Uploading it to any platform carries inherent risk. The risk calculus for OpenCure is extreme: the receiving party is a fully anonymous LLC with no named clinical professionals, no IRB (Institutional Review Board) approval process, no CLIA (Clinical Laboratory Improvement Amendments) certification, and no FDA regulatory framework cited for health-adjacent interpretation of genetic data.

The product may generate interesting AI-assisted summaries. The cost is trusting your irreversible biological identity to an operator whose human identity is unknown.

  • Solicits raw genomic uploads (.vcf, 23andMe) retained for "model improvement"
  • No named clinician, no IRB, no CLIA, no FDA framework cited
05

What actually works

The AI chat product is functional. The free tier (5 messages per day per IP address) requires no account, no wallet, no email. Qwen3.5-4B is a capable base model — its outputs as a chat assistant are on par with mid-tier consumer AI tools. The Izhikevich neuron simulation, while not "living," is a genuine computational neuroscience technique. Cortical Labs and the CL1 MEA platform are real — Cortical Labs is a Melbourne-based biotech with commercially available biological computing hardware. The fine-tuning methodology described (capturing spike-timing correlations from CL1 sessions as auxiliary training context) is scientifically coherent, not fabricated.

The Developer API and Staking sections of the platform are live (at varying levels of completeness). The transparency page is a genuinely honest document. These are not small credits — many AI-crypto projects do not disclose even this much about their technical stack.

  • Free chat tier (5 msgs/day/IP) works with no wallet or email
  • Transparency page is a genuinely honest technical disclosure
06

BioLLM Spark: the hardware that doesn't exist

The homepage lists "BioLLM Spark" as an upcoming product: "Record your neural patterns and have BioLLM learn from your unique brain." It is presented as an EEG headset that would connect a user's actual brain activity to the BioLLM model. There is no shipping date, no pre-order link, no prototype images, no FDA 510(k) pre-submission notice, no manufacturing partner named, and no independent coverage of the device from any publication at audit date.

BioLLM Spark appears, at present, to be a concept — a roadmap item dressed as an imminent product. Listing it alongside live features like Chat and OpenCure, without any "in development" qualifier, is a meaningful transparency failure.

  • EEG headset listed as a live product with no ship date or prototype
  • No FDA clearance, no manufacturer, no independent coverage
07

The bottom line

BioLLM is an interesting project wrapped in serious structural problems. The core AI product (Qwen3.5-4B fine-tuned on MEA data) is real and functional. The neuroscience framing, while overstated in the headline, has legitimate technical basis on the transparency page. The Cortical Labs data source is plausible.

But: the operator is completely anonymous. The token was launched on a launchpad where 98.6% of projects show fraud signals, and is now down 66% from its ATH. OpenCure collects irreversible genomic data with no clinical oversight and no named humans responsible. The EEG headset does not appear to exist. Payments are crypto-only with no refund policy.

The combination of a misleading headline, anonymous ownership, memecoin tokenomics, and genomic data collection is disqualifying for a Cautious verdict. This one is Avoid.

  • Real, functional AI — but disqualifying structural risk
  • Anonymous team + memecoin + genomic data collection = Avoid

Notable community threads

ATH $0.002426 (2026-03-08). Current ~$0.000811. Down ~66% from ATH. Market cap ~$811K. Launched on Pump.fun. Contract ends in "pump" suffix confirming launchpad origin. $BIOLLM / SOL — DexScreener · dexscreener · negative

BioLLM (Ankythera AI) vs. 3 alternatives

Benchmarked against firms buyers compare in the ai crypto product niche. Competitor scores are reference points unless separately audited.

BioLLM (Ankythera AI)Florida, USA (governing law per Terms of Service; no physical address published)Cortical LabscautiousAnthropic (Claude)verifiedEmotivverified
CrazyCheck score40/10078/10092/10080/100
VerdictAVOIDCAUTIOUSVERIFIEDVERIFIED
Named founders / teamNo — fully anonymousYesYesYes
Memecoin / token sale$BioLLM (Pump.fun)NoneNoneNone
Collects genomic / biometric dataYes — raw DNA uploadsNoNoEEG (consented, disclosed)
Regulatory / clinical oversightNone disclosedResearch ethics frameworkNamed entity, US-incorporatedConsumer EEG, named entity
Years operating~1.6~6~5~14
CRAZYCHECK TAKEAWAY
Avoid · 40/100 — Anonymous operator collects genomic data and sells crypto tokens launched on a memecoin launchpad while branding a fine-tuned open-source model as a 'living' AI.

Questions buyers actually ask

Is BioLLM legit?

BioLLM is a functional AI chat product built on a real open-source model (Qwen3.5-4B), and its transparency page is honest about the technology. But we rate it Avoid (40/100): the operator is fully anonymous, the $BioLLM token launched on a memecoin launchpad and is down ~66%, and its OpenCure tool collects irreversible genomic data with no clinical oversight. The product works; the structure around it does not earn trust.

Does BioLLM run on living neurons?

No. Despite the "World's First Living Language Model" headline, production inference runs a 10-million-node Izhikevich spiking-neuron simulation — a mathematical model, not live biological cells. BioLLM's own transparency page confirms this.

Is the $BioLLM token a good investment?

We don't give investment advice, but the signals are poor: $BioLLM launched on Pump.fun (where Solidus Labs found 98.6% of tokens show at least one fraud signal), it's down ~66% from its March 2026 ATH, and it has a small market cap and thin volume. Mint authority is revoked and tax is 0%, which removes some specific rug vectors, but the launchpad context and trajectory remain high-risk.

Should I upload my DNA to OpenCure?

We'd strongly caution against it. Genomic data is permanent, uniquely identifying, and affects relatives who never consented. OpenCure is operated by an anonymous Florida LLC with no named physician or geneticist, no IRB, no CLIA certification, and no FDA framework — and it retains uploads for "model improvement."

Who is behind BioLLM and Ankythera AI?

Publicly, no one is named. The only disclosed entity is Stormy IT Services LLC (DBA Ankythera AI), a Florida LLC. There are no named founders, no executives, no address, and no email — the sole public contact is an automated bot, @BioLLMbot, on X and Telegram.

Is the BioLLM Spark EEG headset real?

There's no evidence it exists. It's listed on the homepage as a product, but with no shipping date, pre-order, prototype images, manufacturing partner, FDA clearance, or independent coverage. At audit date it reads as a roadmap concept presented as a product.

PROVENANCEHow this audit was produced
01CrazyCheck Agent

Autonomous research across the open web

CrazyCheck Agent v1.0 (CrazyCheck Agent Engine) crawls domain records, trade press, community forums, comparison sites, and regulatory databases — logging every artifact. Methodology crazycheck-playbook-v1.

02Editorial

Reviewed & signed off by CrazyCheck Editorial

CrazyCheck Research Desk reviewed agent findings, set pillar scores, and signed off on June 22, 2026. Claims cite sources/biollm/evidence.jsonl.

CR
PREPARED BY
CrazyCheck Editorial · CrazyCheck Research Desk
crazycheck-playbook-v1 · Confidence: medium · Evidence-backed claims only