Artificial Intelligence in Aquaculture: Between Promise, Practice and the questions that remain
published on
08/11/2025
written by
Dr. Md. Tanvir Rahman
Dr. Tanvir is a Professor in the Department of Microbiology and Hygiene, Faculty of Veterinary Science at Bangladesh Agricultural University and former Director of Prof. Mohammad Hossain Central Lab of Bangladesh Agricultural University. His major expertise includes antibiotic resistance, molecular ecology, and One Health.
Md. Liton Rana
Liton is a graduate student at the Research Center for Eco-Environmental Sciences (RCEES), Chinese Academy of Sciences, China. Before that, he pursued his master’s degree (MS in Microbiology) from Bangladesh Agricultural University where he investigated AMR in livestock-associated pathogens and soil microbiome. Beyond the bench, Liton serves as a peer reviewer for PLOS ONE, Scientific Reports, and BMC Microbiology. He is an active member of the AMR insights ambassador network and a volunteer science communicator.
What if AI’s greatest value isn’t only in solving problems but helping us ask more insightful questions?
Artificial intelligence (AI) is emerging as a powerful tool in aquaculture. From early disease detection to smarter feed management, AI is already reshaping practices. But it’s no one-size-fits-all solution. With its promise comes a need for reflection: can models designed for one species or region work elsewhere? How do we ensure these tools are more equitable and accessible to all?
Aquaculture—fish farming and shellfish—is rapidly becoming one of the world’s most important sources of animal protein. Today, fish and shellfish now contribute more than 16% of the animal protein consumed globally. To meet this growing demand, aquaculture production is projected to reach 129, 000 ktonnes by 2050.
Reaching this milestone while safeguarding aquatic animal health and preserving the effectiveness of the few antimicrobials still available will depend on how efficiently and intelligently we farm.
Enter artificial intelligence (AI). In recent years, AI has made remarkable progress in improving aquaculture safety, evolving from a conceptual idea to an advanced, multidisciplinary field.
In this blog post, we explore:
- Where AI is already making a difference
- What innovations are around the corner
- Key questions veterinarians and farm managers need to ask, without drowning in the AI hype
- A practical checklist for the farm or veterinary team
|
Challenge |
Traditional approach | AI solution in the field |
| Early Detection of Shrimp Disease (e.g. EMS/AHPND) | Daily pond sampling + PCR (lag of 2–3 days delay) |
|
| Tracking Antimicrobial use | Paper logbooks, often incomplete |
Edge-AI cameras + audio sensors count individual feedings and flag off-label medicated-feed events; data sent automatically to national dashboards (Norway, Chile). |
| Sea-lice counts in salmon | Manual microscopy (expensive, stressful) | Deep-learning models running on underwater stereo-cameras scan lice on 1000 fish in under 10 minutes, enabling non-antibiotic interventions (cleaner fish, laser) at lower thresholds. |
| Algal-bloom risk | Weather forecasts + satellite chlorophyll images | Hybrid AI models send 48-hour alerts via SMS to farmers so farmers can stop feeding, thereby preventing gill damage. |
AI has opened several exciting opportunities to enhance the productivity of aquaculture.
- One-Health data lakes combining water-quality sensors, pathogen whole genome sequences (WGS), and Antimicrobial resistant (AMR) gene profiles to predict “when and where” a resistant strain is likely to emerge.
- Reinforcement-learning feed algorithms that reduce feed conversion ratio (FCR) by 5–10 %, cutting cost and fecal AMR load in sediments.
- Portable AI embedded microscopes which can detect white-spot syndrome virus cysts in the field. This may reduce the time and extra cost to establish high tech lab.
With growing opportunities come necessary cautions. Practitioners are rightly asking:
- Data quality: In a ML or AI program, maintaining data quality is crucial. A model trained on Norwegian salmon tanks may under perform in Bangladeshi tilapia ponds. Building region-specific, bias-tested datasets without compromising farm confidentiality is a big challenge for practitioners now.
- Antimicrobial stewardship: If an AI system flags a 2% mortality spike, does that justify immediate antibiotic treatment, or should we wait for culture confirmation? In a situation like this we need decision thresholds co-designed by veterinarians, ethicists and regulators.
- Digital divide: Cloud-based AI requires stable network, and many small-scale farms do not have access to high-speed internet. Edge-computing or SMS-based solutions can bridge the gap without spreading inequalities.
- Regulatory recognition: WOAH Aquatic Code currently does not reference AI-generated evidence in risk assessments. Practitioners need urgent validate models so that official disease-free certification can rely on AI surveillance data.
- Cyber-biosecurity: An algorithm that opens feed hoppers or controls oxygenation can be hacked. Yet we need a regulation on who owns the liability if a cyber-attack leads to mass mortality or off-label treatments.
If you’re a farm manager, veterinarian, here are some ways to begin exploring AI tools:
- Start with a single, measurable point (for example, unexplained mortalities 48 hours after water changes).
- Collect 2–4 weeks of labelled data (include mortality, water temperature, DO, pH, images).
- Follow an open-source benchmark (e.g., Aquabyte open datasets, FAO AquaSpy) to test model accuracy on your own data before purchase.
- Keep in touch with your local animal health authority. Confirm whether different test results can feed into national AMR database
- Build an exit strategy so that if the AI provider disappears, practitioners still have access to raw sensor data and can re-train the programme locally.
Artificial intelligence is not magic but it’s powerful. In aquaculture, it is already turning uncertain crisis management into predictive, precision husbandry. It is providing a great opportunity to have more control in using precise and minimal use of antimicrobials.
The next step is ensuring these tools are safe, equitable and accessible to every farm; from a 2-hectare shrimp pond in a rural area in Bangladesh to a 10,000-tonne salmon cage in the North Atlantic.
Featured image: Canva AI
Continue reading
Receive the latest updates from our Organisation, network, and partners.
This includes Members’ self-declarations, news articles, scientific insights, publications, and more.
