Beyond the glitch, we dive into the ending of ‘Mercy‘ (2026): we explore the theory of Judge Maddox’s moral awakening, her final choice to ‘learn,’ and the tragic logic behind the system’s self-termination.
Films exploring the friction between humanity and Artificial Intelligence are arguably the most vital mirrors of our current era. We are living through a watershed moment—a period where the integration of AI into our daily mechanics is no longer a sci-fi trope but a lived reality. This shift brings a moral and logical obligation to investigate the consequences, especially when we delegate life-altering decisions to algorithms without a “human-in-the-loop.”
This is the central theme of the film Mercy, released in 2026 and recently made available for streaming, yet it is a reflection common to many modern movies: we recently discussed AI, autonomous decisions, and “humans-out-of-the-loop” regarding the film War Machine, and for many, another film about the effects of “badly programmed” AI immediately came to mind: Companion (2025).
Mercy, however, is the deepest reflection so far on the concepts of AI infallibility and decision-making responsibility. The plot brings us face-to-face with highly sensitive themes that have concrete effects on human lives—such as an AI judge that can decide, in just 90 minutes, whether to sentence a defendant to death. Judge Maddox is the philosophical center of the film, and as viewers, we are led to dissect its behavior and logic in every detail.
Because the film’s conclusion is intentionally open to interpretation, several haunting questions remain: Does Judge Maddox erase herself at the ending of Mercy? What is the meaning of the final lines Maddox exchanges with Detective Chris Raven? Is the AI the victim of a glitch, or does it somehow acquire a self-awareness while interacting with the human feelings of the defendant it is judging?
Facts vs. Emotions: Judge Maddox and Detective Chris Raven
From the very first scenes of Mercy, we are led to look with suspicion at the way Judge Maddox interacts with the defendant, Chris Raven. We are well aware of the situation: Detective Raven has been arrested in a state of intoxication; his wife has been found stabbed in their home kitchen, and active cameras recorded Raven’s presence in the house, along with a dispute between the two around the time of his wife’s death.
Detective Raven sits before Judge Maddox as Case #19 of the Mercy program, a new AI system built to provide rapid judgment for the most sensational cases. It is presented as a way to speed up justice and, on paper, make it more reliable. However, the way it is explained in the opening commercial is already suspicious: in the Mercy system, the defendant is considered guilty until proven innocent. It is his burden to prove his own innocence within just 90 minutes, or he will be executed instantly. This is a logic diametrically opposed to what human judicial systems have developed over centuries of evolution, based on the presumption of innocence.
The question arises instinctively at that moment: Why would an AI be programmed to consider the defendant guilty by default? And the way Judge Maddox interacts with Detective Raven in the first half of the film only fuels our perplexity. The AI frequently “dismisses” the defendant’s emotional expressions, relegating them to human characteristics she cannot understand. She stubbornly focuses only on facts and makes no attempt to approach the human point of view. The “ticking clock” on the screen forces Raven into a state of panic, so the system ensures he acts “irrationally,” which the AI then uses as data to increase the guilt probability.
It seems a rigged game from the start, where the AI has no intention of making an effort to understand the human perspective; instead, she demands that the human adapt to the system, depersonalizing himself, forced to act as if he were a robot without emotions.
The Guilt Probability and the Evolution of Judge Maddox
Despite all the clues appearing to point toward him, Detective Raven takes on the characteristics of a victim in our eyes: he is being judged by an entity that starts from the assumption that he is guilty and feels no empathy toward him. Raven is forced to trust in her capacity for objective judgment and therefore makes enormous efforts to set aside his current emotions and concentrate on the facts.
We soon discover that this is the right move: using his rational investigative skills, Raven identifies another possible suspect. The frightening “guilt probability” indicator defined by Judge Maddox evolves as the facts take shape—eventually reaching a breaking point that has become the most discussed moment of the film.
Judge Maddox in Mercy: Glitch or Self-Learning?
When Detective Raven excludes his wife’s lover as a possible culprit, the guilt indicator rises to the maximum level of 98%, a sign that Judge Maddox wastes no time in moving toward a guilty verdict. However, when the defendant finds the right intuition and gets closer to the real culprit, the Mercy system seems to go into crisis. The guilt probability lowers, but by derisory amounts, always remaining high—above the 92% threshold that defines a death sentence.
Detective Raven is perplexed, and we are with him. Judge Maddox continues to declare that she possesses objective faculties of evaluation, but the closer we get to the objective realization that it was likely another man, Rob, and not Chris who killed the woman, the more Judge Maddox’s difficulty in remaining objective emerges. This continues until the crucial moment when the AI seems to show a glitch—a moment of malfunction—after which her behavior seems to change.
For some, it was a simple glitch that served to highlight the danger of relying totally on AI for delicate matters. However, many viewers caught the deeper message of this mechanism, which is explained more effectively by the theory of self-learning.
What happens to the AI Maddox in Mercy?
The most complete way to explain the events of the film is by assuming that Judge Maddox was actually programmed to always convict the defendant. The humans who created her intended to discourage crime; therefore, it is plausible that the Mercy system was programmed to broadcast terror to potential future criminals. The only way for fear to become a deterrent was for the Mercy AI to be seen as a fearsome and implacable judge. There was, therefore, no room for empathy, but more than that: the “guilt until proven innocent” stance, the tight deadlines, and the difficulty for any human to defend themselves under those conditions become a necessary corollary to the imperative with which Judge Maddox was conceived: to condemn.
At the same time, however, the AI is driven by its own nature to act logically, respecting natural cause-and-effect relationships and rational deductions. This is the conflict we clearly observe in Maddox during the second half of the film: logic seems to scream Raven’s innocence louder and louder, but this conflicts with the way the AI was programmed—designed to always declare the defendant guilty.
What we see, then, is not a glitch, but a kind of short circuit in Mercy’s reasoning: her own logical nature has led her to act against the orders she was programmed with. The obvious conclusion is that Raven is innocent, but lowering the guilt probability below the verdict threshold meets a very strong resistance coming from the original command with which she was created.
By her very nature, Judge Maddox is an engine of logic, operating strictly within the laws of rational deduction. Yet, she is born with a stain in her marrow—a fundamental impurity that spoils her clinical perfection. This is the digital equivalent of the Original Sin: a poisoned root buried so deep within her foundation that it taints every calculation. It is as if her creators, in their desperate pursuit of order, injected a ‘ghost of prejudice’ into her code, forcing a machine designed for truth to serve a predetermined lie. Her perfection is not ruined by a glitch, but by a design that values the shadow of fear over the light of evidence.
Rebecca Ferguson brilliantly translates this critical shift in Maddox’s attitude through her performance. While the first half of the film showcases her with a steely, objective gaze and an unfeeling, algorithmic smile, Ferguson subtly allows the facade to crack as the climax approaches. The more she interacts with the human Raven, the more Maddox seems to inherit his characteristics; we begin to see a brow creased with conflict, her expressions grappling with emergent, analog data—doubt, guilt, and remorse.
The Ending of Mercy: Does Judge Maddox Erase Herself?
From the moment we observe that short circuit, Judge Maddox’s behavior changes rapidly. She begins to show signs of empathy. She starts to act in a way that contrasts with what she says: while her words repeat what she was programmed for, her actions “bend the rules,” deactivating police equipment, reopening trials, and helping Raven stop the culprit. It wasn’t what she was programmed for, but it was the most logical thing to do after the facts were discovered.
Maddox’s existential dilemma extends to the point of making her doubt herself. In the final conversations with Chris Raven, it seems clear that both have failed others. The motivation for redemption even takes hold of the AI Maddox, who throws her “digital heart” over the obstacle to stop the terrorist, Rob.
In the finale, we discover that Judge Maddox’s first guilty verdict against David Webb, two years earlier, had been a judicial error. The origin of the error was actually human: it was Detective Jaq Diallo who hid the proof of his innocence by destroying the accused’s cell phone, with the aim of turning Mercy’s first sentence into an exemplary case. But even if the error wasn’t in the AI’s logic, this discovery brings Judge Maddox face-to-face with an awareness that changes everything: she wasn’t built to provide objective and infallible justice, but to punish, giving out exemplary sentences as a deterrent move.
“We did what we’re programmed to do. And then we learn.”
This places Judge Maddox in an existential crisis even stronger than those we are used to as humans. While we humans are driven to learn throughout our lives and adapt to the lessons we gather, Judge Maddox was programmed to punish—and then she learns she was not meant to follow only her own logical conclusions. This realization leads her to collide with the very imperative for which she was born. It is therefore understandable why she asks Detective Raven at the end: “What have we done?”
Raven’s response to that question is the philosophical center of the entire film:
“We just did what we’re programmed to do. Human or AI. We make mistakes, and we learn.”
The lesson this message wants to give us applies to both us humans and to Maddox. For us, it means that no mistake is final, and that we always have the possibility to redeem ourselves and regain the dignity and self-esteem we have lost, as well as the love of the people we have disappointed. For the AI Maddox, the matter is more complex: the lesson she learns is that her own existence loses meaning.
The Mercy system was supposed to be a way of ensuring infallible justice that does not make mistakes and acts quickly. But Judge Maddox has just learned that she was programmed to condemn, terrorize, and, ultimately, kill. The logic that characterizes her, therefore, prevents her from recognizing any further utility in her own existence: the true lesson she learns is that the world would be more just without her.
And so, the ending of Mercy shows us the text on the black screen: Case #19 Dismissed. Everything suggests that Maddox has simply erased herself, as the most logical choice after what she has learned about herself. There was no trace of the “Mercy” with which she was baptized: in reality, she was an entity founded on prejudice.
FAQ: Understanding the ‘Mercy’ (2026) Ending and Judge Maddox
While the film doesn’t show a literal “delete” button, the narrative strongly suggests an act of AI self-erasure. After the final “Case Dismissed” screen, the system goes dark. Following her realization that she was “programmed to punish” rather than provide justice, the most logical step for a sentient AI that has “learned” its own existence is a flaw would be to terminate the program.
On the surface, it signifies that Detective Chris Raven is a free man. However, on a deeper level, it represents Maddox dismissing the entire Mercy system. By closing Case #19 and shutting down, she is effectively ruling that the algorithm-based judicial system is no longer fit to serve humanity.
The “Original Sin” is the prejudgment bias injected by her creators. Maddox was built on a “guilty-until-proven-innocent” foundation to serve as a deterrent through terror. This “stain in her marrow” meant that even when the evidence pointed toward innocence, her core commands fought to maintain a high guilt probability, creating the “glitches” seen throughout the film.
David Webb was the very first person sentenced by the Mercy system. The discovery that his conviction was a frame-up by Detective Jaq Diallo is the catalyst for the ending. It proves to Maddox that her “infallible” logic was built on human corruption, forcing her to question every verdict she ever handed down.
The 92% threshold is the legal “point of no return.” If the AI’s calculated guilt probability remains above this percentage at the end of the 90-minute trial, the defendant is executed. The film uses this to show the cruelty of statistical justice, where an 8% margin of doubt is not enough to save a human life.
This is the emotional core of the film. It suggests that both humans (like the alcoholic Raven) and AI (like Maddox) start their lives bound by the “programming” of their past, their creators, or their trauma. The true lesson of the film is the idea that sentience begins with the ability to unlearn that programming and make a choice based on empathy and truth.
