The model is built to detect when crimes are being "contemplated." A US telecom company trained an AI model on years of inmates' phone and video calls and is now piloting that model to scan their calls, texts, and emails in the hope of predicting and preventing crimes. Securus Technologies president Kevin Elder told that the company began building its AI tools in 2023, using its massive database of recorded calls to train AI models to detect criminal activity. It created one model, for example, using seven years of calls made by inmates in the Texas prison system, but it has been working on building other state-or county-specific models. Over the past year, Elder says, Securus has been piloting the AI tools to monitor inmate conversations in real time (the company declined to specify where this is taking place, but its customers include jails holding people awaiting trial, prisons for those serving sentences, and Immigrations and Customs Enforcement detention facilities). "We can point that large language model at an entire treasure trove [of data]," Elder says, "to detect and understand when crimes are being thought about or contemplated, so that you're catching it much earlier in the cycle."

Thailand's top court has ruled that former prime minister Thaksin Shinawatra must serve a year in jail, in yet another blow to the influential political dynasty. The decision relates to a previous case where he was sentenced to years in prison for corruption, but ended up spending less than a day in a jail cell as he was moved to a hospital. On Tuesday, the Supreme Court ruled that this transfer was unlawful - and that the 76-year-old would have to serve his sentence in jail. Thaksin and his family have dominated Thai politics since he was first elected PM in 2001. His daughter Paetongtarn previously served as leader but was removed from office last month over a leaked phone call.

The rapid advancement of multi-modal large reasoning models (MLRMs) -- enhanced versions of multimodal language models (MLLMs) equipped with reasoning capabilities -- has revolutionized diverse applications. However, their safety implications remain underexplored. While prior work has exposed critical vulnerabilities in unimodal reasoning models, MLRMs introduce distinct risks from cross-modal reasoning pathways. This work presents the first systematic safety analysis of MLRMs through large-scale empirical studies comparing MLRMs with their base MLLMs. Our experiments reveal three critical findings: (1) The Reasoning Tax: Acquiring reasoning capabilities catastrophically degrades inherited safety alignment. MLRMs exhibit 37.44% higher jailbreaking success rates than base MLLMs under adversarial attacks. (2) Safety Blind Spots: While safety degradation is pervasive, certain scenarios (e.g., Illegal Activity) suffer 25 times higher attack rates -- far exceeding the average 3.4 times increase, revealing scenario-specific vulnerabilities with alarming cross-model and datasets consistency. (3) Emergent Self-Correction: Despite tight reasoning-answer safety coupling, MLRMs demonstrate nascent self-correction -- 16.9% of jailbroken reasoning steps are overridden by safe answers, hinting at intrinsic safeguards. These findings underscore the urgency of scenario-aware safety auditing and mechanisms to amplify MLRMs' self-correction potential. To catalyze research, we open-source OpenSafeMLRM, the first toolkit for MLRM safety evaluation, providing unified interface for mainstream models, datasets, and jailbreaking methods. Our work calls for immediate efforts to harden reasoning-augmented AI, ensuring its transformative potential aligns with ethical safeguards.

Drones have become a "threat to national security", the prisons watchdog has said, after a surge in the amount of weapons and drugs flown into high-security jails. Charlie Taylor, the chief inspector of prisons, called for urgent action from Whitehall and the police after inquiries found that terrorism suspects and criminal gangs could escape or attack guards because safety had been "seriously compromised". His demands follow inspections at two category A prisons holding some of England and Wales's most dangerous inmates. HMP Manchester and HMP Long Lartin in Worcestershire had thriving illicit economies selling drugs, mobile phones and weapons, and basic anti-drone security measures such as protective netting and CCTV had been allowed to fall into disrepair, inspectors found. In a report released on Tuesday, Taylor said the police and prison service had "in effect ceded the airspace above two high-security prisons to organised crime gangs" despite knowing they were holding "extremely dangerous prisoners".

Joseph Voigt, 23, fled after leaving Marvin Voigt, 63, dead and Susan Voigt, 58, with a gunshot wound to the head. Police responded to the scene after Susan Voigt reported the incident at around 11:20 p.m. on Saturday. The Bartow Police Department said they arrived to find Marvin Voigt dead in the driveway from apparent gunshot wounds and Susan Voigt inside the home suffering from a serious gunshot wound. She was taken to a hospital in critical condition. "They found the mother sitting up on the couch," police chief Stephen Walker told reporters, according to Fox 13. "She was alive. She had been shot in the head once."

Things to Do in L.A. Tap to enable a layout that focuses on the article. Gabriel Zendejas Chavez leaves the federal courthouse in downtown Los Angeles. Indicted in 2018 in an investigation of the Mexican Mafia's rackets in L.A. County jails, Chavez pleaded guilty in federal court to a rarely filed charge called "misprision of a felony." A lawyer accused of helping members of the Mexican Mafia traffic drugs, collect extortion money and expose government informants pleaded guilty Thursday in a deal with prosecutors that may spare him prison time. Gabriel Zendejas Chavez, who was indicted in 2018 in an investigation of the Mexican Mafia's rackets in L.A. County jails, told U.S. District Judge George Wu he was guilty of a rarely filed charge of "misprision of a felony."

It was a eureka moment for Jibran Ilyas. Like much of his party, the Pakistan Tehreek-e-Insaf (PTI), Ilyas had been swamped by a sense of uncertainty. Their charismatic leader, former Prime Minister Imran Khan, has been in jail for months. Senior party officials are in hiding. Campaigning in any meaningful way for the February 8 elections to the National Assembly and provincial legislatures appeared difficult, if not near-impossible.

A faux-rap song that features the AI-generated voice of Donald Trump defending himself against the criminal indictment out of Georgia has jumped to the No. 2 spot on the iTunes rap chart. Hi-Rez the Rapper, who describes himself as a radical freedom extremist, released the song on August 25, one day after Trump posted his August 24th mugshot to X (formerly Twitter). The song has been viewed nearly 3million times on X, and has shot to the top of the Hip-Hop charts since its release. The digital imitation of the former president's voice is nearly identical to the real-life 2024 candidate's. In the song, Hi-Rez imagines how Trump responded to being booked and released on bond on charges of attempting to overturn the 2020 election in Georgia.

RABUPURA, India – Their love affair across one of the world's most heavily guarded borders had begun on the virtual battlefields of a video game, where players bond over having one another's back against bloody enemy ambushes to become the last survivors. But when Seema Ghulam Haider, 27, a married Pakistani Muslim, sneaked into India with her four children to be with Sachin Meena, 22, a Hindu man, their time together was brief. About two months after they started secretly living in the same neighborhood outside New Delhi, the couple ran into difficulties with the Indian authorities. This week, Haider and her children were arrested on charges of having illegally entered India. Meena and his father were also arrested, on charges that amount to little short of conspiring to shelter an enemy.

While the systemic brutality used by Colombian police to quell national protests in 2021 was real and is well documented, photos recently used by Amnesty International to highlight the issue were not. The international human rights advocacy group has come under fire for posting images generated by artificial intelligence in order to promote their reports on social media – and has since removed them. The images, including one of a woman being dragged away by police officers, depict the scenes during protests that swept across Colombia in 2021. But any more than a momentary glance at the images reveals that something is off. The faces of the protesters and police are smoothed-off and warped, giving the image a dystopian aura.