August 11, 10:07

Artificial Intelligence Transforms Cyber Threats: Recovery Becomes as Crucial as Prevention

As software development accelerates due to artificial intelligence, the cybersecurity landscape is changing. Advanced AI models, such as Mythos, are capable of detecting vulnerabilities with unprecedented speed, while autonomous agents, AI-generated code, geopolitical uncertainty, and shrinking attack windows force organizations to rethink how they protect critical systems.

These topics were central to the round table 'Cyber Resilience in the Mythos Era: Moving at Machine Speed in an Uncertain Digital World,' organized by Rubrik and YourStory Media in Bengaluru on July 10, 2026.

A number of specialists from various sectors participated in the discussion, including fintech, e-commerce, logistics, enterprise software, and education. These included Ananth Nag, Vice President APAC Rubrik; Aditya Chandra, VP of Platform Engineering MoEngage; Anshul Sharma, Director of Engineering Scaler; Apoorva Gaurav, Senior VP of Engineering Clear; Arghya Mukherjee, Head of AI and Data Algonomy; Chidambaran Subramanian, CTO InCred Financial; Gaurav Kapatia, Head of Engineering Newton School and Newton School of Technology; Karthikeyan Ramasamy, Senior Director of Engineering Freshworks; Mohan Devarapalli, Head of Engineering PayU; Navin Kumar, VP of Engineering LeadSquared; Niraj Kumar, Director of Engineering Shadowfax; Prabod Goel, Global Engineering Lead Postman; Rishabh Chhajer, VP of IT and Cybersecurity Allen; Utkarsh Tiwari, Head of Security Engineering (Cloud Technologies) Meesho; and Vikas Roy, Director of Engineering Vahan.ai.

Participants discussed the challenges companies face as AI integrates into products and operations. While concerns covered protecting Personally Identifiable Information (PII), securing AI-generated code, managing AI agents, and minimizing supply chain risks, one message dominated: prevention alone is no longer enough.

Since organizations use AI for faster software creation and deployment, they are simultaneously generating new security threats. Participants pointed to prompt injection attacks, AI-generated code, non-human identifiers, third-party dependencies, and AI agent actions on behalf of users as some of the most serious problems enterprises must solve.

Mukherjee emphasized the risks of using customer data to train Large Language Models (LLMs) and the need to secure agentic AI systems. Kumar spoke about protecting sensitive customer information and preventing cross-tenant data leakage in LeadSquared, while Ramasamy noted the emerging challenge of managing non-human AI identities. Others discussed the growing need for more robust AI guardrails, observability, and governance as AI becomes part of daily business operations.

The very speed of software development creates an additional layer of complexity. As Sharma stated, 'AI expands the attack surface faster than we can defend it.' Although participants agreed that AI has accelerated development and experimentation, they also noted that this same speed increases the likelihood of vulnerabilities that organizations may not notice or immediately understand.

Several participants suggested using AI within a defensive strategy. Roy explained how Vahan.ai employs defensive agents and AI-based detection mechanisms, stating: 'We use AI to fight AI.' He acknowledged that the approach is still evolving because AI agents are often non-deterministic, which complicates their testing and protection using traditional methods.

Meesho focused on securing AI-generated code from the outset. Tiwari noted: 'We start by putting guardrails on the coding agent itself so that the code it writes is secure by default.' Nag strongly urged the group to look beyond prevention and consider what happens after a breach. He referenced recent incidents with Jaguar Land Rover and Marks & Spencer, noting that even large organizations with mature security programs can face prolonged downtime. He argued that companies traditionally invested heavily in prevention, detection, and remediation, whereas resilience—the ability to restore business operations—often received less attention.

Participants agreed that the conversation is shifting as cyberattacks are increasingly viewed as inevitable. Instead of asking, 'Can we stop every attack?', organizations are starting to ask, 'How quickly can we recover systems if an attack succeeds?'

Ramasamy countered that while AI brings new risks, organizations must continue to rely on established security principles, including multi-layered architecture, identity-based access, least privilege rights, and robust guardrails. He stressed that AI agents should only access authorized tools and data on behalf of verified users, thereby limiting potential breach damage.

For Chandra, resilience is also linked to organizational culture. He argued that many companies still view cybersecurity as a compliance exercise rather than an engineering discipline built on continuous monitoring, behavioral threat detection, and operational readiness. He added: 'Culturally Indian organizations are not ready. Many businesses do it just for show, for ticking boxes, for an audit checkmark. We do not realize the depth of the problem.'

The discussion repeatedly returned to one conclusion: in an AI-driven world, preventing attacks remains critically important, but resilience holds equal weight. As software development accelerates and attack cycles continue to shrink, organizations will need to prepare not only to defend their systems but also to recover quickly in the event of a breach.

A constant theme was the distinction between disaster recovery and cyber recovery. Resolving operational outages is relatively straightforward because organizations know their backups remain untouched. Participants noted that cyberattacks are fundamentally different: attackers can compromise both production systems and backups, leaving organizations uncertain about which data copies they can trust.

Explaining this difference, Nag said: 'When we are hacked, it is not operational recovery. Operational recovery is my copy that I just need to get back. Cyber recovery is I have a copy, but I don't know if that copy is infected. If it is infected, what do we need to recover?' He emphasized that true cyber resilience depends not only on maintaining redundant infrastructure. Organizations need clean, isolated recovery environments and verified backups that can be restored with confidence after an attack.

One participant shared a case where a disgruntled database administrator intentionally deleted critical databases. Recovery was only possible because the organization had maintained several isolated copies of its data. This experience, participants noted, reinforced the value of continuous, protected backups and clearly defined recovery strategies capable of countering both external and internal threats.

As enterprises deploy more AI agents, participants agreed that organizations should begin treating them like any other digital identifier—with specific permissions, monitoring, and accountability. Prompt injection attacks, wandering agents, and identity management for non-human users became among the biggest challenges as AI systems become more autonomous. Speakers discussed the importance of policy guardrails, adversarial testing, sandboxing, identity management, and continuous monitoring to ensure the safe operation of AI systems. While no single security system can eliminate all risks, participants agreed that multiple layers of defense significantly reduce the probability of unexpected AI behavior.

The discussion also touched upon how AI-generated code is changing software development itself. Since engineering teams are releasing code faster, developers are increasingly implementing software they may not fully understand, making governance, visibility, and continuous verification more important than ever.

By the end of the discussion, participants agreed that cyber resilience extends far beyond technology. Many organizations still view cybersecurity as a compliance requirement rather than a core business opportunity. Building resilience requires regular incident modeling, recovery drills, and tabletop exercises to ensure teams can quickly resume operations in the event of an incident. Participants also stressed that resilience must become a board-level priority, not solely the responsibility of technology and security teams. They argued that recovery planning should be seen as an investment in business continuity, not just another security initiative.

Throughout the round table, one idea kept surfacing: AI accelerates software development, but it also increases the speed and complexity of cyber threats. As organizations integrate AI into products, engineering, and operations, they are also expanding the number of systems, identifiers, and workflows that need protection. For many present, this changes the very nature of cybersecurity. The goal is no longer to assume that every attack can be prevented. Instead, organizations must be prepared to quickly detect attacks, restore trusted data, resume business operations, and minimize downtime when a breach occurs. In a threatening, AI-driven landscape, resilience ceases to be a backup plan; it becomes an integral part of the cybersecurity strategy.

super.abril.com.brAugust 12, 22:00

Exome sequencing test: limitations and the need for complementary approaches in genetic diagnosis

Exome sequencing, a widely promoted genetic test, has faced criticism due to its limitations, despite having represented a revolution in the study of rare diseases since the early 2010s. Initially, it allowed for the simultaneous analysis of thousands of genes, leading doctors and health plans to adopt it as the primary method of clinical genetics, popularizing the idea of a 'complete genetic assessment.'

However, this popularization generated a false belief that a negative result from the exome completely ruled out any genetic cause. Yet, there is a significant logical flaw in this understanding, as the exome examines only coding genes, which account for only 1% to 2% of human DNA.

Experts point out that the causes of serious neurological diseases often reside in the other 98% of genetic material, in non-coding regions. These areas act as regulators of when and how genes should be activated or deactivated. Additionally, pathological alterations can occur in repetitive sequences, in mitochondrial DNA, or through epigenetic modifications, as seen in specific syndromes.

Investigating only a small fraction of the genetic map is insufficient to be considered a comprehensive test. Statistical data demonstrates this restriction: the exome test diagnoses only about 27% of children with developmental delays and only 16% of autism cases. For the vast majority who remain without answers, the investigation is concluded under the label of 'disease of unknown cause.'

A crucial point lies in how medicine communicates uncertainty. When the laboratory reports that 'no alteration was found,' technically this refers to the limited scope of the test. However, many families interpret this phrase as a definitive verdict that their child's problem has no genetic origin, representing a huge discrepancy between technical information and emotional conclusion.

For families who persist without a diagnosis, diagnostic medicine offers alternative paths. Clinical genetics is a dynamic field, with constant discoveries of new genes and associations of mutations with human diseases. One study observed patients with negative exomes five years ago; when their old data was submitted for clinical reevaluation and analysis with modern software, the diagnosis rate jumped from 31% to 53%.

This increase is due to approximately 42% of the new diagnoses arising from recent scientific advancements, genes and variations that were not present in the original databases. Automated reanalysis systems allow old data to be cross-referenced with updated scientific literature in an agile manner. Thus, requesting periodic reanalysis of the exome is a fundamental step for those who have already undergone the test.

When exome reanalysis is inconclusive, the definitive solution is Whole Genome Sequencing. Thanks to technological advances, the cost of mapping 100% of human DNA has drastically decreased, making it comparable to the previous cost of the exome. This method transcends the partial view of the exome by investigating the totality of the genetic code, including structural faults and previously invisible regulatory regions, transforming a test once restricted to academia into an accessible clinical practice.

It is imperative to change the perspective of diagnostic medicine: the exome must be seen as a starting point, not as the final limit of genetic investigation. Reanalyzing old data and transitioning to whole genome sequencing are actions that restore hope and the right to an answer for families.

olhardigital.com.brAugust 12, 20:47

Google Unveils New Pixel Device Lineup, Including Pixel 11 Smartphones, Pixel Watch 5, and Pixel Tag

Google introduced its updated Pixel device series during an event in the United States. The announcement included the Pixel 11, Pixel 11 Pro, Pixel 11 Pro XL, and Pixel 11 Pro Fold smartphones, as well as new gadgets such as the Pixel Watch 5, Pixel Buds Pro 2, and the brand new Pixel Tag.

Furthermore, the company demonstrated several improvements in artificial intelligence, photography, content creation, accessibility, and health monitoring. This strategy expands Gemini's influence within the Google ecosystem, allowing the company to compete in segments already dominated by other manufacturers.

The devices were presented before the official start of sales and will reach consumers at different times. Mobile phones will begin shipping on August 20th, while the Pixel Tag is expected in November.

Details of the New Pixel 11 Smartphones

The base model of the new series now costs $899 USD, which is $100 higher than the Pixel 10. The base configuration has received doubled memory capacity, reaching 256GB, while retaining 12GB of RAM. The device body features a 6.3-inch OLED display with a resolution of 1080 by 2424 pixels and a maximum brightness of 3000 nits. The main changes concern the rear camera, where the protrusion has been reduced compared to the previous generation.

In terms of photography, the Pixel 11 is equipped with a 48MP main camera, a 13MP ultrawide camera, and a 10.8MP telephoto lens with up to 30x zoom capability. The main sensor has increased in size, and the stated battery life exceeds 30 hours. Performance is powered by the Tensor G6 chip, which, according to Google, provides 25% faster internet operation, reduces application launch time by 15%, and improves energy efficiency compared to the Tensor G5.

The company also links the new processor to improvements in image processing and locally executed Gemini functions. Among the innovations are faster transcription and AI-dependent features directly on the device. Two advanced variants start from $1099 and $1299 respectively. Both variants include 256GB of storage and 12GB of RAM; choosing the 512GB or 1TB versions allows for up to 16GB of RAM. The panels of these models reach a brightness of 3600 nits, and Google claims that scratch resistance has doubled compared to the previous version. Batteries have also slightly increased, and the company promises over 30 hours of use.

The camera module has received significant enhancements: Pro Zoom now reaches 120x, exceeding the previous 100x, and the devices have also received updates to their camera sensors. The most notable feature is HiLight—a function that uses an LED ring around the flash to convey information without screen notifications. The system can signal when Gemini is listening, processing a request, or responding, and can also notify about calls from selected contacts, including WhatsApp messages. Google plans to expand the capabilities of this indicator further.

The foldable model is available for $1899 and integrates into the new lineup through structural changes and software updates. The outer screen has increased to 6.5 inches, and both displays can reach a brightness of 3600 nits. Google announced reduced bezels and presented the device as more durable than its predecessor. The gadget has IP68 certification against dust and water, and the back panel is made of a material better resistant to impact. Despite similar dimensions to the previous model, the device has become lighter: weight decreased from 258g to 238g, and battery capacity decreased from 5015mAh to 4750mAh.

The camera system retains most of the configuration of previous models, but the main camera aligns with the changes made to the Pixel 11. Super Zoom also reached 30x. Storage starts from 256GB with options for 512GB and 1TB. The device uses Tensor G6, which, according to Google, provides up to 24 hours of battery life.

Gemini's Central Role in the New Pixel Series

The new generation of Pixel places even greater emphasis on Gemini. Among the new features is an agent capable of performing user tasks, such as ordering goods, booking trips, and making calls to companies. When making calls, users will be able to monitor real-time transcription and answer the conversation, although this feature will initially only be available in the US.

The Gemini Intelligence feature allows retrieving information from authorized applications. This could include interpreting a class schedule in Gmail to determine necessary books or checking the calendar to respond to an invitation. Google is also preparing Rambler, which will be integrated into the keyboard to help eliminate filler words from voice texts. Live Translate will gain support for translating voice messages, videos, and podcasts.

The camera of the new Pixels has received numerous features for everyday shooting and content creation. One of them is Magic Capture, which combines photos and video clips from a single moment, applying automatic cropping and noise reduction. The Camera Looks feature differs from simply applying filters because it alters the image processing itself to create specific visual styles, including the aesthetics of editorial publications and digital cameras.

In the Pro models, video also received stabilization for 8K recording. The new Creator Suite includes tools such as a teleprompter, clip editing, and voice enhancement. The ability to use Circle to Search directly within the camera window has been added, allowing the user to learn information about the frame without leaving the app or taking a pre-shot.

Among the functions presented by Google is a tool that converts signals in front of the camera into text. It uses Google DeepMind's Sign Language-to-Text model and will initially support transcription in American Sign Language English. This technology will be integrated into Gboard and Live Transcribe to facilitate communication between people who use sign language and those who do not, reducing the need for manual text input.

Updates to the Pixel Watch 5 Smartwatch

The new smartwatches start from $399 USD, which is $50 more than the Pixel Watch 4. The device is available in 41mm and 45mm sizes with various finish options, and there is a special edition honoring basketball player Stephen Curry priced at $579. The stated battery life is up to 30 hours for the 41mm model and up to 40 hours for the 45mm model; using power-saving mode, the larger version can last up to 72 hours. Fast charging has also been improved: after 15 minutes of charging on the dock, up to 15 additional hours of use can be gained.

The watch retains the Actua 360 display with Gorilla Glass and a maximum brightness of 3000 nits. Internally, it is equipped with Snapdragon W5 Gen 2 Accelerated, which, according to the manufacturer, provides 20% faster processing than the Pixel Watch 4. Storage capacity has also doubled to 64GB.

In the health sector, the Pixel Watch 5 expands the range of metrics. The watch now tracks blood pressure trends and insulin resistance, and also provides data on sleep breathing quality. Google also introduced the Readiness Score, which helps determine whether the user should focus on recovery or training. The Google Health app will receive personalized strength training and HIIT programs with visual and audio cues for exercises and rest periods.

Another feature is Health Coach, available to Google Health Premium subscribers. Using Gemini, this tool can create workout plans adapted to fitness levels and daily routines, and also offer recommendations for sleep and recovery. The Pixel Watch 5 also supports emergency response and breathing monitoring features. According to company materials, the Breathing Emergency Detection feature can initiate an emergency call upon a sudden and prolonged drop in oxygen saturation level.

Introduction of the Pixel Tag Tracker

The Pixel Tag is Google's first proprietary Bluetooth tracker under the Pixel brand. The device costs $29 USD when purchased separately or $99 USD in a pack of four. Integrated into the Find Hub Android network, this accessory allows users to locate items and emit an audible alert. The ecosystem also enables initiating searches via Pixel Watch and using Gemini commands in Pixel Buds.

The tracker has an elongated shape, measuring 1.1 by 1.8 inches, and weighs 11 grams with the installed battery. The device is equipped with a wideband chip for more accurate location determination. Power is supplied by a replaceable CR2032 battery, estimated to last over a year. The accessory has water and dust resistance, with two materials having different classifications: one indicating IP67, and the other indicating submersion resistance up to one meter for 30 minutes. Release is scheduled for November 11th, and it is compatible with Android 9 or later devices.

The Pixel Buds Pro 2, which were already available on the market, were also mentioned. Google introduced an olive green version that harmonizes with the new devices. In addition to the new look, the buds will receive a firmware update in September, which should improve Active Noise Cancellation (ANC) performance.

In conclusion, the new Pixel smartphone series shows a price increase of $100 compared to predecessors: the Pixel 11 starts at $899, the Pixel 11 Pro starts at $1099, the Pixel 11 Pro XL starts at $1299, and the Pixel 11 Pro Fold starts at $1899. Pre-orders for the smartphones will begin on August 12th, with shipping starting on August 20th. The Pixel Watch 5 is available from $399, and the special Stephen Curry edition is priced at $579. The Pixel Tag will be released later, on November 11th, at a price of $29 per unit or $99 for a set of four.

olhardigital.com.brAugust 12, 20:40

WhatsApp tests username-based contacts and automatic group messages on iOS

WhatsApp has begun implementing two distinct changes for iOS users, focusing on both contact management and group dynamics.

One of the changes allows new contacts to be saved using only a username, eliminating the need to link them to a phone number. This feature is already available to certain users on both Android and iOS, according to WABetaInfo.

This addition aligns with the progressive expansion of the username system within the app. Simultaneously, Meta is testing on iOS a tool designed to send automatic welcome messages to members joining groups.

The new field for registering a username is located in the WhatsApp contact creation screen. To access it, the user must press the ' +' button in the 'Chats' section and select 'New Contact'. Previously, identifying a contact was strictly tied to their phone number.

This improvement also helps optimize privacy features. For example, a contact registered this way may have permission to view certain statuses or profile data, such as the icon and description, even when those options are restricted only to the contact list.

The novelty has started appearing for individuals who updated WhatsApp through the App Store, and beta users have also located this new field. However, there is no official timeline established for its full release, as distribution occurs gradually.

Regarding group functionalities on iOS, Meta is developing an option that will allow administrators to automatically trigger a greeting for each new participant. This function had already been identified in the Android beta version and recently appeared in the iOS beta version 26.31.10.73, although it has not yet been released to testers.

The welcome message will have a specific bubble called 'automatic greeting'. It is important to note that the rest of the group members will not be able to see it, but the new member can interact with the content. Even in groups with multiple administrators, there will only be one automatic welcome message, whose text will be shared among all of them and editable by any administrator.

While the welcome tool is still under development without a set date for the public, the username feature has already started appearing for a portion of users, diversifying the ways of identification within WhatsApp.