Hive AI detector
Vendor Castle Global, Inc. (Deploy) Official site thehive.ai/apis/ai-generated-content-clas... How this card is sourced
Hive AI-Generated Content Detection processes text in chunks of 2,048 characters each, according to the Hive detection docs read 8 November 2026. The vendor is Castle Global, Inc., trading as Deploy, at 100 1st Street, 11th Floor, San Francisco, CA 94105, per the Deploy privacy policy read the same day; a company blog post of 26 October 2026 announced that Hive is now Deploy, so the card keeps the name Hive, now Deploy. Hive is one entry in the index of AI detectors, and this card uses the same fields as the other twenty-one. How this card is sourced.
What Hive detects
| Field | Value | Source |
|---|---|---|
| Content types detected | Text, image, video, voice, music | docs.thehive.ai/docs/ai-generated-content... |
| Generators covered | Text: not published (single ai_generated class), Image and video source head: sora, pika, kling, luma, runway, flux, dalle, stablediffusion, midjourney, imagen, adobefirefly, bingimagecreator, ideogram, gemini, veo3, imagen4, sora2, gptimage1_5, grok and about 90 further classes, Music attribution: duobao, google, heartmula, loudly, minimax, mubert, mureka, musicgen, riffusion, stable_audio_open, suno, udio, yue, ace_step | docs.thehive.ai/docs/ai-image-and-video-d... |
| Input limits | Text: processed in chunks of 2048 characters each; maximum length not published. Image: gif, jpg, png, webp. Video: mp4, webm, avi, mkv, wmv, mov. Audio: flac, mp3, ogg, wav, m4a plus video formats; scored per 10-second chunk (music per 3-second chunk). File size, media length, resolution and codec limits depend on project configuration; contact support. | docs.thehive.ai/docs/supported-file-formats |
| Free tier | Free AI detector Chrome extension (no fees or Hive login required) and Hive Detect web tool; no cap stated. Free API quota for AI detection not published (every AI Detection pricing line is Contact Sales). | thehive.ai/apis/ai-generated-content-clas... |
| API | Yes, docs.thehive.ai/docs/ai-generated-content... | docs.thehive.ai/docs/ai-generated-content... |
| Integrations | Hive AI Detector Chrome extension, X Bot, Hive Detect (hivedetect.ai), Enterprise Application (no-code), Hive moderation dashboard | thehive.ai/apis/ai-generated-content-clas... |
| Data retention | Docs FAQ: Our default policy is to retain customer data for 14 days. We can increase or decrease our retention time upon request according to your needs (including not retaining your data at all). Increasing retention time may incur additional charges. Deploy Privacy Policy (last updated October 23, 2026): We will delete your personal information when it is no longer required for these purposes. | docs.thehive.ai/docs/frequently-asked-que... |
| Ownership | Castle Global, Inc. (Deploy), 100 1st Street, 11th Floor, San Francisco, CA 94105. Brand: Hive is now Deploy (blog, October 26, 2026). | deploy.com/privacy |
| Independent studies | 2 records | Studies index |
Hive covers text, image, video, voice (audio) and music, per the detection docs read 8 November 2026. The product page hero says the APIs "scan images, video, and audio", while the docs and the Chrome extension copy include text; the card follows the docs because they enumerate the endpoints. For text, Hive publishes a single ai_generated class and names no generators.
For image and video, the source head returns classes including sora, pika, kling, luma, runway, flux, dalle, stablediffusion, midjourney, imagen, adobefirefly, bingimagecreator, ideogram, gemini, veo3, imagen4, sora2, gptimage1_5 and grok, with about 90 further classes (image and video detection docs, read 8 November 2026). Music attribution returns duobao, google, heartmula, loudly, minimax, mubert, mureka, musicgen, riffusion, stable_audio_open, suno, udio, yue and ace_step.
On retention, the docs FAQ says: "Our default policy is to retain customer data for 14 days. We can increase or decrease our retention time upon request according to your needs (including not retaining your data at all)" (Hive docs FAQ, read 8 November 2026). The Deploy privacy policy, last updated 23 October 2026, says personal information is deleted "when it is no longer required for these purposes"; the card leads with the 14-day figure because the FAQ states the only concrete period. Integrations are the Hive AI Detector Chrome extension, an X Bot, Hive Detect at hivedetect.ai, a no-code Enterprise Application and the Hive moderation dashboard (product page, read 8 November 2026). Two related cards cover several media types: Pangram for text, image and video, and Copyleaks for text, code, image and video.
Limits and plans
| Plan | Price | Limits | Checked on |
|---|---|---|---|
| AI Image + Deepfake Classifier | Contact Sales (also a higher-limits variant). | ||
| AI Video Detection | Contact Sales (also a higher-limits variant). | ||
| AI Audio Classifier | Contact Sales (also a higher-limits variant). | ||
| AI Music Detection | Contact Sales. No AI text detection line on the pricing page. Tiers: Developers, pay-as-you-go usage billing; Enterprise, custom pricing with highest rate limits; On-Prem (Offline) and Enterprise Application, Contact Sales. |
Text is processed in 2,048-character chunks. Image uploads accept gif, jpg, png and webp; video accepts mp4, webm, avi, mkv, wmv and mov; audio accepts flac, mp3, ogg, wav and m4a plus the video formats, scored per 10-second chunk, with music scored per 3-second chunk (supported file formats, read 8 November 2026). File size, media length, resolution and codec limits depend on project configuration, and Hive says to contact support for them.
The plan rows are from the Deploy pricing page read 8 November 2026, which carries no AI text detection line and no published figure on any AI detection row. The free tier is the Hive AI Detector Chrome extension, which needs no fees or Hive login, and the Hive Detect web tool, with no cap stated on either (product page, read 8 November 2026). The free tier has no stated period.
What the vendor claims
Industry-leading model performance. 'Hive is the clear winner': A 2024 independent research study found that our AI-generated detection model outperforms competing models as well as human expert analysis. Deploy page: Deploy's AI-Detection models outperform competing solutions on internal evaluations, third-party benchmarks and independent studies. (No numeric accuracy figure published.)
Hive says: "Industry-leading model performance. 'Hive is the clear winner': A 2024 independent research study found that our AI-generated detection model outperforms competing models as well as human expert analysis." The Deploy page adds that "Deploy's AI-Detection models outperform competing solutions on internal evaluations, third-party benchmarks and independent studies." Both statements are on the product page (thehive.ai/apis/ai-generated-content-classification, read 8 November 2026), and neither carries a number.
Hive also says: "We recommend that you start off with these thresholds, but you should always refine these thresholds to suit your specific use case", and that the source head returns none when no source can be identified (image and video detection docs, read 8 November 2026). The docs add that Hive APIs do not support batch processing.
What independent studies found
- AIDI-S-0017Ha and Passananti 20242024
Image detector: 98.03% accuracy, 0% false positive rate and 3.17% false negative rate on 280 human artworks and 350 AI-generated images from five generators; detection of AI images fell to 67.56% under high-intensity Glaze while the false positive rate rose to 3.23% (Tables 1 and 4).
- AIDI-S-0030RAID image 20252025
Image detector: 0.60 (accuracy) on 50 clean AI images, 0.44 on the same 50 images after RAID transferable adversarial perturbation, 1.00 on 50 real images, at a 0.5 score threshold; mean AI score fell from 0.59 to 0.45 (Table D.13, labelled preliminary by the authors; no API version or date printed).
No study in the index had tested Hive before 9 November 2026; one record on image detection was added that day, and the detectors with a record are ranked on AI detectors ranked by independent testing.
Hive's product page quotes a 2024 study calling Hive "the clear winner" without naming the study, the sample or a rate; the phrase matches Ha and Passananti 2024, indexed 9 November 2026, which does name all three. A citation a reader cannot follow carries the weight of a slogan, not a result.
One record was added on 9 November 2026. Ha and Passananti 2024 ran Hive's image detector with a paid account on 280 human artworks from 53 artists and 350 AI-generated images from five generators and reports 98.03% accuracy, a 0.00% false positive rate and a 3.17% false negative rate, with detection of AI images falling to 67.56% and false positives rising to 3.23% under high-intensity Glaze; the authors call Hive "the clear winner among all five detectors" (arXiv 2402.03214, read 9 November 2026). The record scored images only; nothing in it bears on Hive's text, video, audio or music detection.
RAID image 2025 joined the list on 9 November 2026 and took the count to 2: Hive scored 0.60 (accuracy) on 50 clean AI images, 0.44 on the same 50 after a transferable adversarial perturbation and 1.00 on 50 real images at a 0.5 threshold, a result the authors label preliminary; the 0.44 entered the robustness field as an image result and the record is counted in the modality records field.
Where Hive appears in the index
Rankings
- AIDI-R-01AI detectors ranked by independent testing Two records added in the fifth wave on 9 November 2026, Pegoraro 2023 and Fleckenstein 2024, changed nine counts and none: Pegoraro 2023 tested ZeroGPT, GPTZero, Copyleaks, Originality.ai, Writer, Content at Scale (BrandWell), DetectGPT, the GPT-2 Output Detector and GLTR, lifting GPTZero to 12, the GPT-2 Output Detector to 10, Copyleaks to 7 and Originality.ai and ZeroGPT to 6; Fleckenstein 2024 tested 289 teachers and no software, so it counts for no card ([Fleckenstein 2024 record](/studies/fleckenstein-2024-teachers/)). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify.
- AIDI-R-06AI detectors ranked by stated data retention
- AIDI-R-07AI image, voice, video and code detectors ranked by independent studies
- AIDI-R-08AI detectors ranked by published robustness to paraphrased and edited text
- AIDI-R-09AI detector false negative rates, ranked
- AIDI-R-11AI detectors ranked by named integration surfaces
- AIDI-R-12AI detectors ranked by the upload formats the vendor states
- AIDI-R-17AI detectors ranked by the generator models the vendor names
Fits these use cases, questions and comparisons
The four rankings order cards by study count, published false positive rate, free words per month and language count. Hive's record carried no value in any of those four fields on the morning of 9 November 2026, so none of the four launch rankings listed Hive; the detector hub lists Hive with the fields it does carry.
Since 9 November 2026 Hive ranks eleventh in independent testing on 1 record, first in studies by content type as the only card with a record on image detection, sixth in false negative rate on 3.17% of 350 AI images, and fifth in data retention on a stated 14-day default.
Hive sits on three audience views: AI detectors for business and AI detectors for copywriters and agencies, because of the API, and AI detectors for job applications, because of the free Chrome extension. Hive also appears on AI detectors by content type, because Hive covers image, video, voice and music alongside text, the widest set of media types in the index. Two single-purpose media cards were added on 9 November 2026 for comparison: Was It AI, an image-only detector with 10 free images a month, and Reality Defender, an enterprise deepfake detector for video, image, audio and text.
Later on 9 November 2026 the RAID image record added the card to the adversarial robustness ranking at 44% on images and lifted its content-type record count to 2.
Hive is one of fifteen cards on the text and image question, added 9 November 2026, with its two image records beside zero text records, and third on the generators covered ranking, added the same day, on 33 named image, video and music classes in its documentation.
Questions about Hive AI detector
What content types does Hive detect?
Text, image, video, voice (audio) and music, per the Hive detection docs read 8 November 2026. Text is scored in 2,048-character chunks, audio per 10-second chunk and music per 3-second chunk, and the image and video source head returns about 90 generator classes.
How much does Hive's AI detection cost?
Every AI detection row on the Deploy pricing page reads Contact Sales, read 8 November 2026, with Developers on pay-as-you-go usage billing and Enterprise on custom pricing. The Hive AI Detector Chrome extension and the Hive Detect web tool are free with no fees or login.
How long does Hive keep submitted content?
14 days by default, per the Hive docs FAQ read 8 November 2026, adjustable on request up or down to zero retention. The Deploy privacy policy, last updated 23 October 2026, says data is deleted when no longer required; the card leads with the 14-day figure.
Is Hive the same company as Deploy?
Yes. The legal entity is Castle Global, Inc., per the Deploy privacy policy read 8 November 2026, and a company blog post of 26 October 2026 announced that Hive is now Deploy. Product pages still carry the Hive name, so the card uses Hive, now Deploy.
Sources and revisions
Sources
- docs.thehive.ai/docs/ai-generated-content...Read 8 November 2026
- thehive.ai/apis/ai-generated-content-clas...Read 8 November 2026
- deploy.com/privacyRead 8 November 2026
- docs.thehive.ai/docs/ai-image-and-video-d...Read 8 November 2026
- docs.thehive.ai/docs/supported-file-formatsRead 8 November 2026
- deploy.com/pricingRead 8 November 2026
- docs.thehive.ai/docs/frequently-asked-que...Read 8 November 2026
Revision log
- Card created
- Ha and passananti 2024 record added, 1 study record
- Raid image 2025 record added in the sixth wave, 2 study records
- Text and image question and generators covered ranking linked (seventh wave)



). Illuminarty, one of four cards added in the same wave, joined with its one image record from Ha and Passananti 2024, so 18 of 72 cards now qualify.: register plate](/img/independently-tested-480.webp)
