PeaceRep research found that amnesties for political crimes are used to remove criminal liability for activities ranging from violent acts in armed conflicts to non-violent political dissent. Most amnesties for political offences are granted in connection with armed conflicts and they are often used in repressive contexts where there are restrictions on political debate and activity (Mallinder 2024).  

Depending on the context, amnesties may benefit a diverse range of opponents of states, including non-state armed groups, coup participants, persons accused of ‘terrorism’ offences in peacetime, political prisoners, leaders and members of opposition groups, and protestors. However, about one-third of amnesties for political offenders also benefit state agents, often where they committed crimes against the state’s political opponents (Mallinder 2024). 

The Amnesties, Conflict and Peace (ACPA) Database (Mallinder and Shaw 2024) contains detailed information on 318 amnesties granted in 94 countries from January 1990 to November 2023. This data shows that the amnesties are sequenced in different ways with peace agreements.  

Of these amnesties, 44 percent were granted during ongoing conflicts and were unconnected to peace negotiations. The absence of negotiations means that these amnesties are generally granted unilaterally by the state through legislation, executive orders, and/or policy decisions. These amnesties are often intended to weaken an insurgency by encouraging combatants to surrender. 

Six percent of the amnesties in the ACPA database were granted as part of a commitment to begin negotiations on a comprehensive peace agreement and were intended to create the conditions for the negotiations to proceed through for example removing obstacles to non-state armed groups participating. For example: 

1. Angola’s Lei de Amnistia (2002) was enacted, after a ceasefire agreement, to implement the government’s peace plan. The plan also contained a number of measures ‘to ensure the necessary legal and political guarantees for the promotion and fulfilment of the process of the national reconciliation’.  
2. Chad’s Loi N0007/PCMT/2021 Portant Amnistie Générale pour des faits d’atteinte à l’intégrité de l’État et aux délits d’opinion was granted to fulfil a series of pre-conditions set by a non-state armed group for entering into peace negotiations. Where non-state combatants are fearful that attending negotiations with government forces could result in their arrest or execution, these amnesties can help to build their confidence that it is safe to attend. However, they can pose political risks for the government, through appearing to give concessions to rebels before they have agreed to permanently end their insurrection. 

Where comprehensive peace agreements are negotiated through a series of partial agreements on specific themes, such as ceasefires, victims’ rights or the return of displaced persons, amnesties may be granted to implement a partial agreement before the comprehensive negotiations are concluded.  

The Peace Agreement Amnesties dataset (Mallinder 2026a, forthcoming), created using data from the PA-X database, codes amnesty provisions from 120 peace agreements from 1990-2024. This data shows that it is rare for pre-negotiation agreements to contain amnesty provisions, but ten percent of partial agreements contain amnesty commitments. Amnesties may also be granted during peace negotiations through domestic law and policy, without being included in a peace agreement. Fourteen percent of the amnesties in the ACPA dataset were granted while negotiations were ongoing. Mid-negotiation amnesties are often confidence building measures. 

23 percent of the amnesties in the ACPA database resulted from comprehensive peace agreements. Indeed, the Peace Agreement Amnesties dataset finds that amnesties have been used in almost one-third of comprehensive peace agreements since 1990. Overall, of the 106 states for which version 9 of the PA-X database hold peace agreement texts, the peace agreement amnesty dataset finds that 47 have had at least one peace agreement with amnesty provisions. This suggests that almost half of all states that have reached peace agreements since 1990 have used some form of amnesty for conflict-related offences (Mallinder 2026b, forthcoming). 

Finally, 12 percent of amnesties in the ACPA database were granted following a comprehensive peace agreement. Often these amnesties were not explicitly envisaged in the agreement itself. They may reverse previously agreed positions about accountability for serious violations or they may be an effort to put into law what had previously been tacit understanding among the negotiating parties that accountability would not be pursued.