Talking Is Not Surrender

Research note: This is a research-grounded normative essay, not a peer-reviewed paper, a systematic review or a probabilistic forecast. The accompanying dossier distinguishes established facts, attributed reports, dated assessments, analytical inferences and publication holds. Election results and September incidents are time-sensitive. Recommendations are the author’s arguments; the sources do not establish that any single programme can guarantee peace. Evidence cutoff: 21 September 2026.

The reason for writing

I do not normally write about party politics. My work concerns technology, infrastructure, sovereignty, and the conditions under which people retain control over their decisions. It has become harder to draw a clean boundary between that work and the state of the world.

Some of what I see frightens me. That is an admission, not evidence.

I have no credible basis for announcing that a third world war is inevitable. I do not believe anyone should manufacture certainty about a catastrophe to make an argument more compelling. But waiting for certainty before taking prevention seriously would be an extraordinary way to manage a risk this large.

The question is not whether we can predict the date of disaster. It is whether we are maintaining the institutions and habits that allow us to avoid one: reliable information, accountable power, credible defence, political participation, and communication with people we profoundly distrust.

My argument is that freedom requires more than the ability to defeat an opponent. It requires arrangements under which opponents can continue to exist without having to destroy one another.

That is not an appeal to become less serious about aggression or extremism. It is an appeal to become more serious about what preventing them actually requires.

1. A warning is not a prophecy

Switzerland’s security strategy of 18 September 2026 explicitly recognizes the risk of the Ukraine war widening into a Russia–NATO conflict. That deserves attention. It does not say such a war is inevitable.¹

The distinction matters. A government should prepare for a dangerous contingency before that contingency becomes probable, let alone certain. Otherwise preparedness would begin at the moment it had ceased to be useful.

Other evidence is troubling. UCDP records 65 state-based armed conflicts in 2025, the highest annual count in its series beginning in 1946. This category is broader than interstate war.² SIPRI estimates global military expenditure at approximately $2.89 trillion in 2025; Europe’s total rose 14%, including spending by Russia and Ukraine.³

Those numbers establish a serious security environment. They do not establish a single coordinated march toward world war. A defence budget can reflect preparations for aggression, fear of aggression, overdue replenishment, or several motivations at once. Expenditure alone cannot identify which explanation applies.

The more specific warning concerns the safeguards around military power. SIPRI reports that New START expired in February 2026 without a successor and describes weakening nuclear transparency and communication.⁴

My concern is therefore not simply that weapons exist, or that governments are buying more of them. It is the possible combination of greater destructive capacity with fewer reliable opportunities to clarify intentions, check assumptions, and reverse a mistake.

That is an analytical concern, not a calculated probability of world war. We should keep the distinction visible precisely because the consequences are so serious.

2. Read the paragraph that interrupts your fear

A frightening interpretation becomes more credible when we test it against the strongest evidence that does not fit.

The same Swiss strategy that warns about Russia assesses that Beijing probably does not currently seek a military attack, given the risks and economic interdependence.¹ The US intelligence community’s March 2026 assessment likewise states that Chinese leaders do not currently plan a Taiwan invasion in 2027 and have no fixed unification timetable.⁵

These are assessments, not guarantees. They do not exclude coercion, a blockade, a later decision to attack, or intelligence failure. They do require us to distinguish a capability, a possible opportunity, and a demonstrated decision.

“China could exploit American distraction” is a scenario worth examining. “China is waiting for its moment to invade” presents an inference as knowledge of intent. The first belongs in responsible risk analysis. The second needs evidence we do not presently have in this research.

India and Pakistan illustrate a related problem: a naval collision on 15 September prompted mutual accusations and protests.⁶ The incident is concerning without an additional unsupported story that both states are waiting for other powers to become occupied elsewhere.

Material constraints also deserve precision. Reuters reported severe pressure on particular US missile stocks during the Iran war, relying on unnamed sources; that is not a public audit of all American ammunition.⁷

Nor are all missiles interchangeable. The European COMLOG contract concerns Patriot GEM-T air-defence interceptors, not weapons designed for deep ground strikes into Russia.⁸

This is not pedantry. A warning assembled from the wrong weapons, inferred intentions, and worst-case assumptions can feel comprehensive while becoming less reliable with every added detail.

There is a plausible mechanism by which separate crises could interact: competition for scarce capabilities, conflicting commitments, misread signals, and leaders’ beliefs about one another’s willingness to respond. But the mechanism must be examined link by link. Several wars do not automatically become one war merely because they appear on the same screen.

3. Talking to an enemy is not accepting his terms

Russia’s war against Ukraine is already a real war between states. Foreign assistance does not reduce Ukrainians to supporting actors in somebody else’s conflict.¹

A call for diplomacy must begin there. It cannot mean that people outside Ukraine decide what Ukrainians should surrender so that everyone else can feel safer. Nor should an appeal to peace turn the aggressor’s choices into the victim’s responsibility.

At the same time, communicating with Russian authorities is not the same as endorsing those authorities, trusting their promises, lifting sanctions, or abandoning accountability. A channel can exist to prevent an incident from escalating even when a settlement remains impossible.

The United Nations Charter accommodates both duties: Article 33 addresses peaceful settlement of dangerous disputes; Article 51 recognizes individual and collective self-defence.⁹¹⁰ Treating defence and diplomacy as mutually exclusive moral identities is therefore a false choice.

The practical questions are harder. Who is authorized to speak? What can each side actually commit to? How will compliance be checked? What happens if an agreement is violated? Which concessions would expose other people to unacceptable danger?

There are talks worth rejecting: a fabricated process used solely to demand capitulation, a negotiation that excludes the people whose lives it determines, or a public spectacle without a credible counterpart. Keeping a capacity to communicate does not require accepting every invitation or treating every demand as reasonable.

But an enemy does not have to become morally acceptable before a conversation becomes strategically necessary.

The Cuban Missile Crisis offers a limited but important example. Its resolution involved communications, restraint and reciprocal arrangements, including the later removal of US missiles from Turkey. The direct hotline followed the crisis; it was not the instrument that had already made the crisis safe.¹¹

The lesson is not that goodwill always prevails. It is that military strength does not remove the need for communication and an executable route away from confrontation.

Even the claim that nobody is talking anymore is too sweeping. On 21 September, China announced a US visit by Xi Jinping for 23–25 September.¹² An announced meeting is no guarantee of progress. It is evidence that diplomacy has not simply disappeared.

We should demand diplomacy that works, rather than pronounce it dead and make its abandonment easier.

4. Democracy owes people rights, not parties power

Inside a democracy, a similar confusion arises when several different activities are compressed into the word “dialogue”.

Listening to a voter is not entering a coalition. Debating an elected representative is not endorsing the representative’s programme. Negotiating an individual measure is not handing over control of the police, the courts, or the machinery of an election. A constitutional party-ban proceeding is different again.

The German debate about a Brandmauer becomes less useful when these distinctions disappear.

My position is that no citizen should be written off as beyond conversation because of a ballot cast. A democratic society needs ways for people to explain their grievances, encounter disagreement, revise a view, and participate without being humiliated.

That does not create an obligation for another party to enter a coalition with their preferred party. A vote establishes representation under the electoral rules; it does not entitle a party to partners willing to govern with it.

We should also reject the claim that history proves exclusion always creates extremism, or that accommodation always moderates it. Conservative elites helped bring Hitler to the chancellorship in 1933, believing they could control him.¹³ This is not a claim that today is identical to 1933. It is a decisive reason not to present inclusion in power as a universal antidote to anti-democratic politics.

The opposite universal claim—that refusing cooperation always protects democracy—would also require evidence, not a slogan. We should assess specific arrangements by their consequences for rights, institutions and political incentives.

There is a defensible boundary: people remain entitled to equal protection and political participation, while institutions must resist attempts to abolish those same entitlements for others.

That principle applies regardless of whether an attack on democracy describes itself as patriotic, socialist, religious, revolutionary, or necessary to save the nation.

It also binds the defenders of democracy. A government cannot make its own conduct democratic merely by identifying an anti-democratic enemy. Its powers must remain reviewable, its accusations contestable, and its opponents protected by law.

5. Election night does not suspend the rules of evidence

September’s German elections deserve serious scrutiny. Provisional results put the AfD first in Saxony-Anhalt with 43.8% and Mecklenburg-Vorpommern with 38.2%. Berlin’s official provisional result puts Die Linke first with 25.7%.¹⁴¹⁵¹⁶

These are different electorates and different results. They are not a referendum establishing what “the average German” thinks about every policy. Being the largest party is not the same thing as winning an absolute majority or already forming a government.

Nor does an alarming election give us permission to lower the standard for allegations. A photograph can establish presence in a room. On its own, it cannot establish a secret bargain, organizational membership, an exchange of votes, or knowledge on the part of every person nearby.

The proper response to a serious allegation is to investigate the conduct, seek corroboration, include the response, and distinguish what is established from what is alleged. A denial does not settle the matter. Neither does an insinuation.

The same discipline applies to policy. Germany’s Basic Law expressly contains an Article 15 permitting specified forms of socialization by law, with compensation provisions.¹⁷ That does not make a particular housing proposal lawful in every detail, affordable, effective, or wise. It does mean that describing socialization itself as proof of fascism substitutes a label for an argument.

Criticize the financing. Examine compensation, legal authority, housing supply, administrative capacity and the effects on tenants. Challenge a policing proposal with its actual budget, powers, safeguards and likely consequences. A politician’s ethnic ancestry answers none of those questions.

Antisemitism must be confronted, including when it appears among people whose other causes we support. Threats against Jews, collective blame, and the celebration of attacks on civilians do not become acceptable when presented as solidarity.

Equally, defending Palestinian rights does not by itself establish antisemitism or support for terrorism. The safety and equal dignity of Jews and Palestinians are not competing principles from which a humanist must select one.

We cannot call for an end to political dehumanization while using it to describe the people we want to reach.

6. Conversation is necessary. It is not magic.

The case for dialogue becomes weaker when we exaggerate what dialogue can do.

A 2024 study tested 25 interventions with 32,059 US participants. Reducing partisan dislike and reducing anti-democratic attitudes were distinct outcomes; different approaches performed best on different measures.¹⁸

Two earlier randomized experiments found that cross-partisan conversations could reduce dislike under some conditions, but the effects faded and did not reliably translate into improved democratic accountability.¹⁹

These findings do not settle the German coalition debate or tell us how to negotiate with a nuclear-armed state. They do warn against assuming that a pleasant encounter repairs an institution, or that simply putting opponents in a room necessarily produces a durable democratic benefit.

My conclusion is narrower: conversation needs a purpose, a structure, a boundary against intimidation, and a way to establish whether it achieved anything.

A neighbourhood discussion about school safety is different from a televised confrontation designed to reward the most aggressive performance. A confidential incident-prevention channel is different from an unverified promise. A listening exercise that never changes a decision is different from meaningful participation.

The Belfast Agreement provides a useful institutional example. Its signatories paired accommodation of substantial political differences with an explicit commitment to exclusively peaceful and democratic means.²⁰

The point is not that this model can simply be exported everywhere. It is that dialogue and enforceable boundaries need not be alternatives. The objective is to make disagreement possible without allowing intimidation to determine the result.

A democratic conversation should offer people a route to change their minds without demanding that everyone else abandon their rights.

7. Capability is moving faster than our arrangements for restraint

Artificial intelligence brings this question into the field in which I work most directly.

The International AI Safety Report 2026 identifies growing cyber misuse and potential assistance with biological and chemical threats, while emphasizing uncertainty about real-world capability and the gap between evaluations and effective harmful execution.²¹

That is serious. It is not evidence that any novice with an open model can reliably create an effective weapon of mass destruction. Information access is only one part of the problem; successful real-world execution is a separate claim requiring separate evidence.

The UK National Cyber Security Centre’s June 2026 warning calls for urgent adaptation to AI-driven cyber risk.²² Here too, a credible warning does not establish that autonomous attackers can already collapse an entire country’s infrastructure at will.

The responsible position is neither dismissal nor theatrical certainty. We should protect systems before harm proves a forecast correct, while refusing to portray every capability demonstration as an operational catastrophe.

For the institutions deploying these systems, I would make the requirements concrete: restricted permissions, separation of critical functions, tested recovery, accountable authorization, and records sufficient to reconstruct what happened. Irreversible, high-impact actions should not become unreviewable merely because an agent can execute them quickly.

The same applies to the political response. A security programme should not use an uncertain threat as a blank cheque for permanent surveillance, unchecked emergency powers, or the removal of legitimate scientific access.

Openness has benefits; AI also has defensive uses.²¹²² The question is where a particular capability, access arrangement or deployment creates an unacceptable risk, and what proportionate safeguards can reduce it without destroying the freedoms supposedly being protected.

A society that responds to every danger by concentrating unaccountable power may become harder to attack while becoming easier to dominate.

8. A humanist position is not the midpoint between two camps

I am not arguing that all governments are equally dangerous or that every conflict has equally responsible sides. Applying one standard consistently does not require distributing blame equally.

The useful questions concern conduct and constraints: can leaders lose power peacefully? Can judges rule against them? Can journalists investigate them? Can minorities seek protection? Can an accusation be tested before punishment is imposed?

Comparative work such as V-Dem’s 2026 report examines precisely these institutional dimensions and their deterioration; its classifications are research assessments, not substitutes for examining particular conduct.²³

Calling every disliked leader “extreme” tells us less than identifying which restraints have been weakened, who has been harmed, and what lawful remedies remain.

A humanist position is not the average of two party programmes. It is a commitment to equal human worth and to institutions that prevent anyone’s certainty from becoming unlimited authority over other people.

That commitment can require resistance. It can justify prosecution for violence, protection against intimidation, sanctions, and credible defence. It can also require negotiating with someone one considers responsible for terrible acts. The relevant test is whether the action protects people and their rights—not whether it demonstrates sufficient hostility to the other camp.

I would therefore ask public institutions for five practical commitments.

Maintain channels before they are needed. Responsible officials should have authenticated, tested ways to clarify dangerous incidents with adversaries. Communication capacity should not depend entirely on a good personal relationship between leaders.

Make defence and diplomacy answerable to a public purpose. Governments should explain what they are protecting, what would count as success, and which diplomatic options remain available. Legitimate operational secrecy should not become an exemption from democratic oversight.

Keep political participation open while protecting constitutional rules. Hear grievances without legitimizing threats. Investigate crime without collective suspicion. Require the same respect for elections, courts, lawful opposition and minority rights from every prospective holder of power.

Invest in recovery as well as punishment. Public safety should include institutions able to withstand disruption and restore essential services. Citizens need evidence that protection works in practice, not only declarations about enemies.

Preserve the possibility of correction. Important decisions should have review mechanisms, identifiable responsibility and conditions for revision. Admitting that an assumption was wrong must remain possible before defending that assumption becomes more important than the people affected by it.

These are proposals, not a scientific guarantee of peace. Their purpose is to preserve the capacity to prevent avoidable harm without abandoning freedom in the process.

9. We still have decisions to make

The most useful question is not whether our time would have sounded unbelievable to somebody living decades ago. It is what we are prepared to do, now, with the choices that remain.

Will we distinguish preparation from prediction? Will we examine an allegation before repeating it? Will we defend an opponent’s rights when doing so brings us no political reward? Will we retain the ability to speak to an enemy without pretending that the enemy has become a friend?

None of this requires optimism about human nature. It requires discipline about human consequences.

The Russell–Einstein Manifesto of 1955 ends with an appeal more useful than a prediction about the weapons of a future war:

“Remember your humanity, and forget the rest.”²⁴

Remembering humanity does not mean forgetting aggression, abandoning accountability, or expecting every dispute to dissolve in conversation. It means refusing to let the identity of an adversary erase the lives affected by our response.

We do not need to agree about everything to recognize that the people who come after us deserve more than the wreckage of our certainty.

Keep the boundaries of democracy firm. Keep the channels of communication open.

Sources