Exactly one week ago, I wrote that Brazil’s institutions had successfully transferred power from Jair Bolsonaro to Luiz Inácio Lula da Silva without requiring Bolsonaro’s personal approval. Lula had won the election, the result had been certified, the outgoing president had left the country, and the constitutional machinery continued operating without him.
Today, thousands of Bolsonaro supporters attacked the physical institutions through which that constitutional order operates.
Crowds in Brasília broke through security barriers and invaded the National Congress, Supreme Federal Court and Palácio do Planalto, the presidential workplace. Windows were smashed, offices entered and public property damaged. The images inevitably recall the attack on the United States Capitol two years ago, although the institutional circumstances are not identical. Brazil’s presidential transition had already occurred: Lula was inaugurated on 1 January and was not inside the Planalto when protesters entered it. What was attacked today was therefore not the mechanism physically certifying Lula’s presidency but the institutions representing the political order that had produced it. Brazilian government documents describe the events as violent invasions of the buildings housing the country’s three branches of government.
Lula responded by declaring a federal intervention in public security in the Federal District until 31 January, arguing that serious failures in local security had allowed the attacks to occur. The intervention is specifically directed towards public security rather than replacing the entire government of Brasília.
There will be time to investigate how the security failure happened, who organised or financed the mobilisation, whether officials facilitated it and what responsibility particular political actors bear. As of tonight, those questions should remain questions rather than conclusions.
The political mechanism visible today is already clear enough.
An election can end without the electoral conflict ending.
This sounds obvious, but democratic systems often behave as though counting the votes completes the political process. Electoral authorities announce a winner, courts resolve challenges, the new president takes office and the system moves on. Legally, that is enough.
Sociologically, it may not be.
Millions of citizens still need to incorporate defeat into their understanding of politics. They need to move from believing “my candidate should win” to accepting “my candidate did not win, so I will oppose the government within the rules until I have another opportunity”.
That psychological transition is one of democracy’s least visible institutions.
There is no government agency responsible for producing it.
It depends on political leaders, parties, media, civil society and citizens themselves reinforcing the idea that losing power is temporary while the constitutional system is permanent.
Brazil’s problem is that a section of Bolsonaro’s movement appears not to have made that transition.
For weeks after the election, supporters maintained protest camps outside military installations, and some openly demanded intervention by the armed forces. The argument was essentially that the institutions responsible for verifying Lula’s victory could not themselves be trusted, meaning another institution, the military, should override the result.
That is not ordinary opposition.
There is an important distinction between believing an election contained irregularities and believing an unelected institution should therefore decide who governs.
Every large election contains mistakes. Citizens have the right to challenge procedures, request audits, bring cases to courts and demand reforms.
Those mechanisms are part of democracy.
The democratic line is crossed when evidence tested through those mechanisms no longer matters because the desired conclusion has already been determined.
If electoral authorities confirm the result, they are corrupt.
If courts reject challenges, the courts are corrupt.
If Congress recognises the winner, Congress is corrupt.
Eventually every institution capable of disproving the allegation becomes incorporated into the allegation itself.
That creates an epistemically closed political movement.
There is no possible institutional answer capable of changing the belief because any institution giving the wrong answer loses legitimacy by definition.
At that point, constitutional politics becomes extraordinarily difficult.
Democracy needs institutions capable of producing authoritative answers to factual questions even when citizens dislike those answers.
How many votes did each candidate receive?
Did the candidate satisfy constitutional eligibility requirements?
Was a particular law violated?
Courts and electoral institutions can make mistakes, and their decisions should remain open to criticism. But political systems cannot function if every citizen treats institutional conclusions as optional whenever another source tells them something more emotionally satisfying.
The alternative is not popular sovereignty.
It is competing realities.
And competing realities eventually produce competing claims to legitimate power.
Brazil is confronting that danger today.
Bolsonaro received 49.10% of valid votes in October. That is not a marginal political constituency. More than 58 million people voted for him. Most of those voters are not storming government buildings, and it would be both analytically wrong and politically dangerous to treat them as though they are.
This distinction may now be more important than ever.
Bolsonaro voters, Bolsonaro supporters and people attacking democratic institutions are not interchangeable categories.
A democracy that responds to extremism by delegitimising an entire opposition electorate creates another problem. Conservative Brazilians need a constitutional route through which they can oppose Lula without being treated as enemies of the state.
The political objective should therefore be to isolate anti-democratic behaviour rather than isolate the entire Brazilian right.
This is strategically important as well as normatively correct.
If ordinary Bolsonaro voters believe the state now considers them collectively responsible for today’s violence, people who currently reject political violence may become more receptive to claims that the new government is persecuting them.
If instead the state distinguishes clearly between lawful opposition and attempts to overturn constitutional government, the extremist coalition becomes narrower.
Good democratic enforcement draws boundaries around behaviour.
It does not criminalise identity.
This creates an enormous responsibility for Lula.
He has every justification for responding firmly to attacks on Congress, the Supreme Court and the presidential palace. The state cannot allow groups to discover that invading constitutional institutions is an effective mechanism for changing political outcomes.
Deterrence matters.
If today’s participants face no meaningful consequences, the expected cost of repeating the behaviour falls.
Political violence can become normalised through impunity.
But accountability must remain individualised and lawful.
Who entered restricted buildings?
Who destroyed property?
Who assaulted officials?
Who organised transportation?
Who financed activities?
Did any public officials deliberately facilitate the attack?
Those are questions for evidence and investigation.
They should not become excuses for broad repression of political opponents.
This is where democracies are tested after attacks against democracy.
Authoritarian governments face a relatively simple response: identify opposition, suppress it and describe suppression as restoration of order.
Democracies have to do something harder.
They have to defend themselves using procedures that preserve the democratic rights they are defending.
That means arrests require legal grounds.
Trials require evidence.
Punishment requires proportionality.
Political speech remains protected even when the speech is offensive or hostile to the government.
Opposition parties remain legitimate even when individuals associated with their broader movement commit crimes.
Democratic self-defence becomes self-defeating if it reproduces the arbitrary power it claims to oppose.
This is the same principle I applied to Peru last month. Castillo’s unconstitutional attempt to dissolve Congress did not give the succeeding government unlimited authority over protesters. Likewise, an attack by Bolsonaro supporters does not give Lula’s government unlimited authority over Bolsonaro’s entire political constituency.
Constitutionalism means accepting constraints precisely when violating them would be convenient.
Brazil’s security failure also requires scrutiny because today’s events should not have been unimaginable. Bolsonaro supporters had been mobilising for weeks, including around military facilities. The possibility of unrest in Brasília was foreseeable.
That raises a state-capacity question.
Democratic institutions need more than legitimacy.
They need physical capacity to protect themselves.
A constitution cannot defend a building.
Police do.
But the police themselves must be institutionally loyal to the constitution rather than to particular political movements.
This is why politicisation of security institutions is dangerous even when no coup occurs.
If officers sympathise so strongly with one political movement that enforcement becomes selective, formal democratic rules begin operating differently depending on who challenges them.
The state loses impartiality.
The investigation into today’s security failures should therefore examine not merely operational incompetence but whether ideological sympathies affected preparedness or response.
Again, that conclusion cannot be assumed before evidence is gathered.
But it is a legitimate institutional question.
The military’s position remains especially important. Some Bolsonaro supporters have spent weeks appealing explicitly for military intervention. That places the armed forces in a situation they should never want to occupy: citizens are asking soldiers to decide a political dispute already resolved by votes.
The correct institutional response is refusal.
A professional democratic military is not neutral between constitutional government and a demand that it overturn constitutional government.
Its neutrality concerns political parties.
Its loyalty concerns the constitution.
Those are different things.
This distinction helps explain why saying the military should “stay out of politics” can sometimes be too vague. If a president illegally orders soldiers to cancel an election, refusing the order is not political neutrality in the sense of indifference.
It is constitutional professionalism.
The armed forces’ democratic role is to remain subordinate to lawful civilian authority while refusing to become an alternative source of political legitimacy.
Brazil’s twentieth-century history makes this especially consequential. The military dictatorship from 1964 to 1985 means demands for military intervention are not abstract fantasies disconnected from national experience.
Institutions remember even when individual citizens do not.
This is why democratic consolidation takes generations.
Rules can change quickly.
Expectations change slowly.
Brazil adopted its current democratic constitution in 1988, but authoritarian political traditions do not disappear simply because a new constitutional document is promulgated. Political movements can continue imagining the military as an emergency political referee long after the formal system rejects that role.
Democratic culture therefore requires repeated practice.
The military refuses political intervention.
Courts resolve disputes.
Election losers accept outcomes.
Governments tolerate opposition.
Another election occurs.
Each repetition strengthens expectations about what normal politics looks like.
Today’s attack attempts to interrupt that normalisation.
The response should strengthen it.
There is another uncomfortable question: how much responsibility does political rhetoric carry for actions committed by supporters?
This is difficult because political leaders cannot automatically be held responsible for everything done by people who support them.
Millions of citizens hear the same speech and behave differently.
Individual agency remains real.
But political rhetoric changes perceived probabilities and legitimacy.
If leaders repeatedly tell supporters that an election system is fraudulent, that institutions are corrupt and that the country is being stolen, some supporters may reasonably infer that ordinary political procedures are no longer adequate.
The stronger the rhetoric becomes, the harder it is to be surprised when someone eventually acts according to its implications.
This does not establish legal responsibility.
It establishes political responsibility for the informational environment leaders create.
Words matter because citizens use them to construct models of reality.
If a president says an election has been stolen, supporters assume the president possesses information unavailable to them.
Authority increases the credibility of the claim.
This is why leaders carry greater responsibility for factual assertions than ordinary citizens.
They possess larger audiences and greater informational authority.
Again, the January 6 comparison becomes relevant, but the broader mechanism matters more than the analogy.
Election denial can function as a coordination technology.
One citizen privately believing an election was fraudulent has little political consequence.
Millions believing it simultaneously, and knowing that millions of others believe it, creates a movement capable of collective action.
Political leaders and social media provide the common knowledge required for coordination.
That can produce peaceful protest.
It can also produce what Brazil experienced today.
This is why democratic parties need internal boundaries.
A political movement cannot outsource all responsibility for extremism to individual supporters while continuing to distribute the narratives from which extremism draws legitimacy.
Mainstream conservative leaders now face a particularly important choice.
They can condemn today’s attacks clearly while continuing to oppose Lula aggressively.
Or they can minimise the violence because condemning it might alienate parts of Bolsonaro’s base.
The first helps institutionalise the Brazilian right.
The second makes its future increasingly dependent on anti-system politics.
This is a collective-action problem inside political parties. Individual politicians may fear losing supporters if they confront extremist narratives. Collectively, the party benefits from preserving a democratic system in which it can eventually return to power.
Short-term electoral incentives can therefore undermine long-term institutional interests.
Leadership means occasionally absorbing the short-term cost.
Lula faces the mirror-image temptation.
He can use today’s events to argue that Bolsonarismo as a whole is incompatible with democracy, strengthening his own coalition through a common enemy.
That might provide short-term political benefits.
It would be dangerous for Brazilian democracy.
A healthy democratic order needs a legitimate right and a legitimate left capable of replacing one another.
If either side believes the other should never be allowed to govern, every election becomes existential.
That increases the incentive to reject defeat.
Democracy therefore needs what might be called mutual political mortality.
I must accept that you can remove me.
You must accept that I can later remove you.
Neither side possesses permanent ownership of the state.
That expectation is what makes peaceful alternation possible.
Brazil’s October election demonstrated electoral competition.
Today’s attack demonstrates that alternation has not yet been accepted universally as legitimate.
This is why last week’s peaceful inauguration was not the end of the democratic test.
It was one stage.
A democracy can transfer formal power while political legitimacy remains contested underneath.
That does not make the transfer meaningless.
It tells us democratic consolidation operates at several levels.
First, rules determine the winner.
Then institutions transfer authority.
Then political actors accept opposition roles.
Then citizens update expectations about the next contest.
Brazil successfully completed the first two.
A section of its political movement is resisting the third.
The state now needs to enforce the boundary firmly enough that the fourth remains possible.
There is a final symbolism in the buildings attacked today. Congress represents elected legislative authority. The Supreme Court represents constitutional and judicial authority. The Planalto represents executive authority.
All three branches were targeted.
That makes the attack more than vandalism against government property.
It reflects a worldview in which the entire institutional system has become illegitimate because it produced the wrong political outcome.
That is the endpoint of institutional nihilism.
If every branch of government is corrupt, every electoral authority fraudulent and every legal process invalid, then only force remains capable of resolving political conflict.
That is precisely what constitutional democracy exists to prevent.
Brazil should therefore respond to today’s attack seriously.
Not because government buildings are sacred.
Buildings can be repaired.
The dangerous idea is that political defeat authorises citizens to bypass the institutions inside them.
That idea has to lose.
Not through collective punishment.
Not through censorship of ordinary opposition.
Not through transforming Lula into the owner of Brazilian democracy.
It has to lose through arrests based on evidence, transparent investigations, independent courts, professional security institutions and conservative politicians making clear that opposition belongs inside the constitutional system.
That response will be slower and less emotionally satisfying than simply declaring everyone involved an enemy.
It will also be more democratic.
One week ago, Brazil demonstrated that institutions did not need Bolsonaro’s permission to transfer power.
Today, it is learning the second half of that lesson.
Institutions also need enough authority to make political defeat final for this election without making it permanent for the people who lost it.
Bolsonaro’s supporters must be allowed to try to win Brazil again.
They cannot be allowed to overturn the Brazil that just voted against them.
That distinction is where democratic opposition ends and democratic destruction begins.
