'This pistol is something a bit different,' according to a firearms expert. Despite its generic appearance, this 18th century firearm features a comparatively unique design. Breakthroughs, discoveries, and DIY tips sent every weekday. It's difficult to resist raising an eyebrow at an Oklahoma robbery suspect's alleged recent weapon-of-choice . According to several Oklahoma news outlets including WKTUL, a 24-year-old man was arrested on December 6 by Tulsa police after allegedly robbing a liquor store using what employees described as an "old-timey musket."

Taki Allen said law enforcement made him get on his knees, handcuffed and searched him, finding nothing. Taki Allen said law enforcement made him get on his knees, handcuffed and searched him, finding nothing. An artificial intelligence system (AI) apparently mistook a high school student's bag of Doritos for a firearm and called local police to tell them the pupil was armed. Taki Allen was sitting with friends on Monday night outside Kenwood high school in Baltimore and eating a snack when police officers with guns approached him. "At first, I didn't know where they were going until they started walking toward me with guns, talking about, 'Get on the ground,' and I was like, 'What?'"

As large language models (LLMs) become increasingly integrated into real-world applications, scalable and rigorous safety evaluation is essential. This paper introduces Aymara AI, a programmatic platform for generating and administering customized, policy-grounded safety evaluations. Aymara AI transforms natural-language safety policies into adversarial prompts and scores model responses using an AI-based rater validated against human judgments. We demonstrate its capabilities through the Aymara LLM Risk and Responsibility Matrix, which evaluates 20 commercially available LLMs across 10 real-world safety domains. Results reveal wide performance disparities, with mean safety scores ranging from 86.2% to 52.4%. While models performed well in well-established safety domains such as Misinformation (mean = 95.7%), they consistently failed in more complex or underspecified domains, notably Privacy & Impersonation (mean = 24.3%). Analyses of Variance confirmed that safety scores differed significantly across both models and domains (p < .05). These findings underscore the inconsistent and context-dependent nature of LLM safety and highlight the need for scalable, customizable tools like Aymara AI to support responsible AI development and oversight.

Recent research has demonstrated that state-of-the-art LLMs and defenses remain susceptible to multi-turn jailbreak attacks. These attacks require only closed-box model access and are often easy to perform manually, posing a significant threat to the safe and secure deployment of LLM-based systems. We study the effectiveness of the Crescendo multi-turn jailbreak at the level of intermediate model representations and find that safety-aligned LMs often represent Crescendo responses as more benign than harmful, especially as the number of conversation turns increases. Our analysis indicates that at each turn, Crescendo prompts tend to keep model outputs in a "benign" region of representation space, effectively tricking the model into fulfilling harmful requests. Further, our results help explain why single-turn jailbreak defenses like circuit breakers are generally ineffective against multi-turn attacks, motivating the development of mitigations that address this generalization gap.

The escalating rates of gun-related violence and mass shootings represent a significant threat to public safety. Timely and accurate information for law enforcement agencies is crucial in mitigating these incidents. Current commercial gunshot detection systems, while effective, often come with prohibitive costs. This research explores a cost-effective alternative by leveraging acoustic analysis of gunshot recordings, potentially obtainable from ubiquitous devices like cell phones, to not only detect gunshots but also classify the type of firearm used. This paper details a study on deciphering gun type hierarchies using a curated dataset of 3459 recordings. We investigate the fundamental acoustic characteristics of gunshots, including muzzle blasts and shockwaves, which vary based on firearm type, ammunition, and shooting direction. We propose and evaluate machine learning frameworks, including Support Vector Machines (SVMs) as a baseline and a more advanced Convolutional Neural Network (CNN) architecture for joint gunshot detection and gun type classification. Results indicate that our deep learning approach achieves a mean average precision (mAP) of 0.58 on clean labeled data, outperforming the SVM baseline (mAP 0.39). Challenges related to data quality, environmental noise, and the generalization capabilities when using noisy web-sourced data (mAP 0.35) are also discussed. The long-term vision is to develop a highly accurate, real-time system deployable on common recording devices, significantly reducing detection costs and providing critical intelligence to first responders.

A new bill moving through the Florida Senate would give homeowners the right to use "reasonable force" to take down drones infringing on their right to privacy, directly conflicting with federal airspace regulations while raising new legal questions regarding how far a person can go to defend their home from surveillance. The bill primarily focuses on further regulating the use of unmanned aircraft systems (UAS) while broadening the scope of locations that are protected from drone flights within the state, such as airports and correctional facilities. Notably, the bill would permit homeowners to use "reasonable force" to stop a drone from infringing on their expectation of privacy. A bill proposed in the Florida Senate would allow homeowners to use "reasonable force" to take down drones infringing on their right to privacy. "No one wants to have a drone sitting over their property, filming what they do for any number of reasons," Florida-based attorney Raul Gastesi told Fox News Digital.

We used contrastive neural networks to learn useful similarity scores between the 144 cartridge casings in the NBIDE dataset, under the common-but-unknown source paradigm. The common-but-unknown source problem is a problem archetype in forensics where the question is whether two objects share a common source (e.g. were two cartridge casings fired from the same firearm). Similarity scores are often used to interpret evidence under this paradigm. We directly compared our results to a state-of-the-art algorithm, Congruent Matching Cells (CMC). When trained on the E3 dataset of 2967 cartridge casings, contrastive learning achieved an ROC AUC of 0.892. The CMC algorithm achieved 0.867. We also conducted an ablation study where we varied the neural network architecture; specifically, the network's width or depth. The ablation study showed that contrastive network performance results are somewhat robust to the network architecture. This work was in part motivated by the use of similarity scores attained via contrastive learning for standard evidence interpretation methods such as score-based likelihood ratios.

Amid a rising tide of low-cost weaponized adversary drones menacing American troops abroad, the US military is pulling out all the stops to protect its forces from the ever-present threat of death from above. But between expensive munitions, futuristic but complicated directed energy weapons, and its own growing drone arsenal, the Pentagon is increasingly eyeing an elegantly simple solution to its growing drone problem: reinventing the gun. At the Technology Readiness Experimentation (T-REX) event in August, the US Defense Department tested an artificial intelligence-enabled autonomous robotic gun system developed by fledgling defense contractor Allen Control Systems dubbed the "Bullfrog." Consisting of a 7.62-mm M240 machine gun mounted on a specially designed rotating turret outfitted with an electro-optical sensor, proprietary AI, and computer vision software, the Bullfrog was designed to deliver small arms fire on drone targets with far more precision than the average US service member can achieve with a standard-issue weapon like the M4 carbine or next-generation XM7 rifle. Indeed, footage of the Bullfrog in action published by ACS shows the truck-mounted system locking onto small drones and knocking them out of the sky with just a few shots.

Los Angeles will utilize AI-powered scanners at Union Station over the next month in an effort to stop passengers with hidden weapons from boarding the rails. Commuters descending to underground platforms for the A, B and D lines (formally known as the Blue, Red and Purple lines) will enter into the testing ground for Metro's 30-day pilot program, which went into effect on Tuesday, though the scanners will not run every day. The program arrives amid growing concern over passenger safety, with Metro recording an uptick in arrests this year for riders carrying concealed weapons. The roughly 6-foot-tall Evolv Technology scanners use artificial intelligence to pinpoint on a person's body where they could possibly be carrying a weapon, according to the company's website. All weapons are banned on the Metro system, and it is illegal to carry a concealed firearm without a permit in California.

A Lancaster man was indicted Wednesday by a federal grand jury on charges stemming from his alleged use of a drone to deliver fentanyl and other narcotics to buyers, one of whom died of an overdose. Christopher Patrick "Crany" Laney, 34, has been charged with one count of distributing fentanyl resulting in death, four counts of operating an unregistered aircraft in furtherance of a felony narcotics crime, one count of possessing methamphetamine with intent to distribute, two counts of possessing fentanyl with intent to distribute, and one count of possessing firearms in furtherance of a drug trafficking crime, according to the grand jury indictment. Federal prosecutors alleged that on several occasions in December 2022 and January 2023, Laney used an unregistered drone to transport fentanyl and other narcotics from his home to a nearby church parking lot, where someone collected the drugs before distributing them to buyers. At least one of those people included a woman who died of an overdose in January 2023. The federal grand indictment also accuses Laney of being in possession of methamphetamine and fentanyl at his home, along with multiple firearms lacking serial numbers -- weapons that are referred to as "ghost guns."