Environment and climate – implementation and finance

That is the judgement awaiting us. Not over our intentions. Not over our conferences. Not over our ceremonial speeches. But over what we actually did when the forests burned, the seas grew warmer, species disappeared and emissions continued.

Sonja Kovalevsky said that it is impossible to be a mathematician without being a poet in one’s soul. In environmental politics, the opposite holds with equal force: it is impossible to love nature without wanting to count it. Love without measurement remains pious words. Measurement without love becomes cold bookkeeping.

We refuse to choose between warmth and precision. We refuse to choose between vision and evidence. We refuse to choose between a society that feels responsible and one that can account for results. That is our entire politics, gathered into a single sentence.

We live in the first age to possess tools large enough to see the Earth as a whole. A satellite sees the methane plume above the landfill before anyone smells it. A model hears the chainsaw in the forest and distinguishes it from the wind. An algorithm knows in advance when the wind will turn and can save the kilowatt-hour that would otherwise have been lost.

This is not science fiction. It is technology already in operation in other countries. What Sweden lacks is not the machinery. It is the political will to direct it. It is the courage to set requirements. It is the determination to say that technology funded by society must also serve society — and prove that it does.

Technology must not become yet another area where profits accrue to a few while the costs are passed on to everyone else. It must be subject to democracy. Not the other way round.

What we know, and what we refuse to pretend

Artificial intelligence changes nothing until it touches something physical. A satellite image must lead to an inspector. A forecast must change a production plan. A sensor must trigger an alarm that someone responds to. A model must lead to an action that cuts emissions, protects the forest, saves water or strengthens society’s preparedness.

Analyses that stop at a report have never saved a watercourse. We will therefore fund the entire chain — data, decision and action — or nothing at all. We will not pay for glossy demonstrations with no power to change reality.

Artificial intelligence is not free. Data centres consume electricity, water, land and metal. Their electricity use increased by seventeen per cent during 2025. An environmental party that turns a blind eye to technology’s own bill has already forfeited the right to demand an account from anyone else.

You cannot preach restraint to households while giving the technology industry a free pass. You cannot demand that citizens measure every kilowatt-hour and then let the biggest systems hide their own consumption behind words such as innovation and the future.

That is why we measure net benefit: avoided emissions and saved resources, minus the system’s own consumption of electricity, water and materials, minus the increased consumption that greater efficiency may unlock. No other figure is good enough as proof.

A promise of thirty per cent is meaningless without a baseline to count from. A press release is not a result. A forecast is not a climate benefit. And a system that cannot withstand scrutiny has no place in public environmental policy.

This is what we will do

A national mission programme for AI, the environment and community development

We start with the problem, never with the technology. It is not society’s job to find excuses for new systems. It is technology’s job to solve real problems in society.

We will fund data, expertise, models, procurement and evaluation as one coherent whole. No municipality should be left alone with a pilot project that works at a trade fair but falls apart in everyday life. At least six pilots must be operational before 2028. Not six reports. Not six vision documents. Six functioning operations that can be measured, scrutinised and expanded.

A Swedish environmental data space

Not a giant database. Not another centralised monument to technocracy. We will build common standards, catalogues and interfaces so that data held by authorities and municipalities can actually be used together.

Today, the knowledge often already exists, but it is locked away in separate systems, divided between authorities or buried in formats that cannot communicate with one another. This is not a lack of information. It is a lack of political coordination. Open by default. Protected where it must be. Shared data must be democratic infrastructure — not a private raw material for whoever can afford to buy access.

See the whole country, every month

Sweden’s forests, wetlands, coasts and habitats must not be inspected only once the damage has already been done. Satellites, laser data, audio sensors, wildlife cameras and field visits must be brought together in a unified system for forests, wetlands, coasts and biodiversity.

The machine identifies where people should look more closely. It never passes judgement. Technology should strengthen human judgement, not replace due process. But responsibility for the decision must always be carried by a human being.

Plug the methane leaks

Landfills, sewage, biogas, agriculture and energy infrastructure must be monitored from space and verified on the ground. Alert. Deadline. Action. Public reporting. That is how simple accountability should be.

Those who emit must not be able to hide behind the fact that nobody happened to measure it. Those who receive an alert must not be able to leave it unanswered in yet another case-management system. And anyone who remedies an emission must be able to show that it really stopped. Methane is the fastest climate gain we have left.

See the fire before it appears

Camera networks, lightning data, satellites and local models must be connected directly to municipal rescue services. In California, the system detects more than one in three fires before the first emergency call.

Minutes determine whether a forest survives the summer. Minutes determine whether a fire can be put out by one crew or requires the efforts of thousands of people. Minutes determine whether families can remain in their homes or are forced to flee. Climate adaptation must not be a document in a folder. It must exist in the alert chain, in the rescue services and in people’s everyday lives.

Get the figures right for wind, solar and heat

We will use better forecasts and managed flexibility in electricity grids, district heating and buildings. Renewable energy that cannot be put to use is not merely a technical failure. It is a political failure.

Every renewable kilowatt-hour we throw away because we could not predict it is a betrayal of the climate, electricity customers and all the people paying for the transition. Society’s job is to make sustainability easy, affordable and accessible — not to moralise at individuals while systemic failures remain.

Water that does not run away

Municipalities will be given tools for optimised pumping, early leak detection and better preparedness for cloudbursts. Water is not a given. Not even in Sweden.

When pipes leak, it is not only water that disappears. Energy, money and future security run away too. Singapore cut its daily production planning from two hours to a quarter of an hour and reduced energy use at the same time. We have bigger leaks and colder winters — and more to gain. That is why every municipality should be able to move from reactive repairs to preventive responsibility.

Close the loops for real

We will procure verified material recycling, not the number of robots. The public sector must not pay for technological props. It must pay for results.

Construction and demolition waste, plastic and textiles will be sorted using machine vision — but payment will be made only once the recycled material has demonstrably replaced virgin material. Otherwise, the loop has not been closed. The waste has merely been moved, sorted or given a new name.

Five safeguards we will never abandon

Not a krona without a baseline

Every project must be measured against a pre-registered starting point and reviewed by an independent party. What does not work will be shut down. And the result will be published even then. Anyone who never reports a failure is probably not pursuing brave innovation policy. They are doing advertising.

An energy and water budget for every public AI system

Small models where small models are enough. Renewable electricity is no alibi for waste. It is not responsible to build a system that saves one kilowatt-hour if it consumes two itself. The whole bill must be brought into the open.

Always a human in the chain

An alert is a reason to investigate. It is never proof. It is never, on its own, grounds for a sanction. Due process is what makes environmental policy legitimate, durable and worthy of people’s trust. The day we allow a formula to assign guilt, we will have created a system in which no one can be held accountable any longer.

Protection for the vulnerable

Cameras and microphones in nature must not become a surveillance network over people. Nor is the nesting site of a threatened species open data. Nature conservation must never become a pretext for mass surveillance. Digitalisation must never become a back door to controlling people’s movements, daily lives and civic associations.

Voters decide the goals, not a formula hidden away

Whoever chooses what a system should optimise is also choosing what society prioritises. That is politics. Should a forest be valued according to timber production, biodiversity, carbon storage, outdoor recreation or local jobs? Should the electricity grid optimise for the lowest cost tomorrow or the greatest resilience over twenty years?

Such trade-offs must never be hidden in code. They must take place in open debate. Conflicts of interest must be reported. Decisions must be open to challenge. Power must be removable. That is the difference between democratic technology and technocratic rule.

Sweden, Europe, the planet

Sweden cannot save the planet alone. Anyone claiming otherwise is engaging in national self-satisfaction, not environmental politics. But a small, technically skilled and trusting country can show the world something it currently lacks: a democracy that directs its technology instead of being directed by it.

We will share data and standards with Europe. We will build joint monitoring of the Baltic Sea together with our neighbours. We will press for EU money to go to systems that measure their own benefit, report their own consumption and can be scrutinised by citizens.

Europe must show that the green transition need not mean democracy handing power over to a handful of global technology companies. That is how a good example becomes an exportable product. Not through boasting. But by making it work, making it open to scrutiny and making it usable by others.

And our solidarity must not stop at Europe’s borders. The countries that have contributed least to the climate crisis are often hit first and hardest. Technology, knowledge and environmental data must therefore be shared on terms that do not create new dependencies and new monopolies. One planet is not just another market. It is the shared condition of our lives.

We are not asking you to trust machines. We are asking you to vote for the people who intend to set requirements for them. People who demand proof. Who report failures. Who refuse to let efficiency become a euphemism for concentrating power. Who protect what cannot speak for itself.

Ellen Key built her life’s work on the idea that the future is something we shape, not something that happens to us. Kovalevsky showed that a conviction becomes strong only when it can be tested. We take both into the Riksdag: love of nature and the demand for proof. The vision and the baseline. The poem and the calculation.

Should technology be used to protect life — or should life once again be subordinated to those who own the technology? Should the transition become a shared reform for freedom — or yet another market where a few take the profits and the many carry the risks? Should we leave our children a living world — or a record of everything we could have done?

Politics is about wanting. But desire must become action. Action must produce results. And results must be tested against reality. The planet is not waiting for better wording. It demands decisions.

Implementation plan

The first 100 days

The Swedish Environmental Protection Agency will be given coordinating responsibility together with, among others, SMHI, the Swedish Civil Contingencies Agency, the Swedish Agency for Marine and Water Management, the Swedish Energy Agency, Lantmäteriet, municipalities and relevant companies. Common measurement rules, baselines and requirements for independent review will be established.

2027

A Swedish environmental data space will open with common standards, catalogues and APIs. It will not be a single central giant database. Open data will be made open, while sensitive environmental, security and personal data will be protected. Six pilot projects will be procured through open problem tenders, rather than the state choosing a supplier in advance.

By 2028 at the latest

• Monthly analysis of satellite and sensor data for forests, wetlands, water and changing land use.
• Methane monitoring with alerts, field checks, an action deadline and public follow-up.
• Early detection of forest fires directly linked to the rescue services.
• Forecasting and optimisation in electricity grids, district heating and larger buildings.
• Leak detection, pump optimisation and cloudburst preparedness in water networks.
• AI-based sorting and verified material recycling.

2029

Independent evaluation will determine which systems should be scaled up, redesigned or stopped. A model that works technically but delivers no measurable environmental benefit must not continue to receive funding.

2030

Approved systems will transition into ordinary operations. By then, Sweden should be able to carry out monthly national reviews of satellite-observable environmental changes, supplemented by field checks.

2031–2034

Effective solutions can be expanded across the Nordic countries, the Baltic Sea region and the EU. Open standards will counter vendor lock-in. The programme is an implementation tool for Sweden’s climate targets, not a substitute for energy, transport, nature conservation or tax policy.

Costs

The following is aipartiet.ai’s planning framework, not supplier quotations received:

The average is no more than 1.25 billion kronor per year over four years. The first authorised funding framework should be 1.3 billion kronor for 2027. The remaining 3.7 billion will be released annually once results have been reported.

The local research includes a concrete but limited cost example: a European demonstration project for robotic sorting received 488 000 euros in EIT funding and approximately the same amount in co-financing, giving a total of around one million euros. This shows the order of magnitude for a single demonstration but cannot be used as the price of a national system. EIT project example

Major investments in, for example, water pipes, electricity grids, recycling facilities or rescue-service vehicles are not included. AI can show where physical investment is needed, but it cannot replace it.

Revenue and funding

There are no certain direct government revenues to book.

The state will fund no more than 70 per cent of each operational pilot. A state pilot framework of 1.2 billion kronor could thereby trigger at least approximately 0.51 billion kronor in co-financing, giving a total pilot volume of around 1.71 billion.

EU funding will be sought from sources including LIFE. LIFE’s 2026 calls total 601.5 million euros across the EU, but nothing may be counted as Swedish funding until a grant decision exists. European Commission LIFE call

Savings in energy, water and materials will primarily accrue to municipalities, authorities and companies. They must therefore not be booked as government revenues.

Environmental taxes, penalty charges and future fines must not be used as uncertain funding items.

Unused funds, particularly the scaling-up reserve, will remain in the Treasury.

The state’s maximum net cost is therefore 5.0 billion kronor over the parliamentary term, less any EU grants actually awarded.

What we know

Sweden’s target is net-zero emissions by 2045 at the latest, followed by negative emissions. In June 2026, the Swedish Environmental Protection Agency assessed that current policy instruments are insufficient. Emissions in 2025 were estimated at 46.7 million tonnes of carbon dioxide equivalent, while the estimated shortfall against the EU emissions budget up to 2030 amounts to 13.1 million tonnes. Swedish Environmental Protection Agency report, Sweden’s climate targets

There are functioning international examples. UNEP’s AI-supported methane system has analysed more than 1.3 million satellite observations, increased analytical capacity 12–15-fold and contributed to more than 40 documented mitigation measures. Human experts verify every alert. UNEP results

Google has reported up to 40 per cent lower energy use for cooling in a data centre. The result is promising but reported by the company and cannot be transferred to other facilities without Swedish testing. Google DeepMind

AI also has an environmental cost of its own. According to the IEA, global data-centre electricity use increased by 17 per cent during 2025, while AI-focused data centres increased by around 50 per cent. Every Swedish project must therefore report electricity, water, materials and total climate impact. IEA analysis

What we do not know

We do not yet know:

• how large the actual emissions reductions or environmental benefits delivered by the Swedish systems will be,
• how many false alarms will arise under Swedish weather and environmental conditions,
• the full cost of sensors, integration, maintenance and field staff,
• how large the models’ energy, water and material costs will be,
• which EU grants will be awarded,
• or which pilots will pass an independent cost–benefit analysis.

That is why no future savings will be counted in advance.

Decide now

The Riksdag will allocate 1.3 billion kronor for 2027. From the outset, every project must report its baseline, total cost, energy use, water, materials, margin of error, responsible human being and the physical chain from alert to action.

If, after no more than 18 months, a project cannot demonstrate a measurable additional environmental benefit, it must be stopped or redesigned. An AI alert must never, on its own, form the basis for a sanction. The 1.5 billion kronor scaling-up reserve may be opened only after independent review. This is Bertrand Russell’s and Sonja Kovalevsky’s precision combined with Jan Stenbeck’s determination to act: we will know what we know, say what we do not know – and start work nonetheless.