Anthropic will begin identifying materials created by Claude, covering texts, images, and other media files, using digital watermarks and metadata that are imperceptible to the human eye.
Anthropic will begin identifying materials created by Claude, covering texts, images, and other media files, using digital watermarks and metadata that are imperceptible to the human eye.
This change is a direct response to the transparency guidelines established by the European Union's Artificial Intelligence Regulation. The standards came into effect on August 2nd and mandate the adoption of clear mechanisms to signal when content has been generated by artificial intelligence.
Although the requirement is European, Anthropic plans to apply this modification globally, covering various access points to Claude, such as APIs, commercial applications, and cloud interactions with partners like AWS, Google Cloud, and Microsoft Foundry.
The company will define two distinct identification methods depending on the type of material generated. For texts, Claude will start incorporating a watermark into the structure of responses, following a model similar to the SynthID used by Google in its Gemini-generated images.
These marks are not visible in the final content, unlike the Gemini logo in Google images or the 'Meta AI' text in Mark Zuckerberg's company system edits. Although Anthropic has not named its mark, it should not modify the content and must remain linked to the text even after copying and transferring it to another software.
For images and visual resources, identification will be carried out through digitally signed provenance metadata, utilizing the C2PA standard, an open protocol already employed by companies such as Adobe, OpenAI, and Google.
It is relevant to mention that Anthropic does not yet have native models for image creation but can produce text files in formats such as PNG, JPG, or SVG.
The application of these measures begins with models launched from the date the European obligations were activated, meaning systems made available after August 2nd must have active identification.
For models and products developed before this deadline, Anthropic stated that it is still working to implement the necessary support. According to The Verge, European legislation grants a four-month grace period for this type of adjustment.
Additionally, the company announced its intention to provide technical documentation and proprietary tools, allowing third parties and online platforms to verify the origin of content generated by Claude in the future.
However, Anthropic warns that these AI identification systems are not infallible. For example, C2PA standard metadata can be lost when sent to certain social networks or converted to other formats. Anthropic itself warns about the possibility of the watermark being lost if the content is rewritten, translated, or if it is excessively brief.
Another notable point is that most platforms do not significantly highlight AI usage; YouTube and Instagram, for instance, insert only a small indication of possible use in discreet areas of their interfaces, and this applies only to images and videos.
Researchers conducted one of the first independent stress tests for GlobalBuildingAtlas (GBA)—a large-scale new digital dataset aimed at mapping every building on the planet. The study, focusing on India's rapidly growing urban landscapes, found that while GBA demonstrates high accuracy in determining building locations on the ground, it systematically mismeasures their height, often underestimating the actual height of skyscrapers by more than half.
This finding is critically important for urban planning specialists and climatologists who rely on three-dimensional models to predict parameters such as energy consumption or floodwater movement in a city.
The work was carried out by a team from the National Remote Sensing Centre (NRSC) at ISRO. Researchers examined various urban areas, including Mumbai, Bangalore, and Hyderabad. As India experiences unprecedented urbanization, having a digital map of its cities has become a necessity for effective resource management.
GlobalBuildingAtlas was recently released as an open-access resource containing over 2.75 billion building markers worldwide. It is the first such resource to offer three-dimensional data at this scale, providing not only building outlines but also approximate heights and simplified 3D shapes. However, since these maps are created using artificial intelligence and satellite imagery, their accuracy can vary significantly depending on local conditions.
To verify GBA, researchers used the NASA ICESat-2 satellite as a space standard. This spacecraft employs a laser altimeter that sends trillions of photons to the Earth's surface, measuring their return time to determine height with centimeter accuracy. The team compared these laser measurements with GBA's height estimates for 20 representative structures, ranging from low-rise industrial warehouses to tall residential towers.
To assess horizontal accuracy, or outline accuracy, the team used high-resolution imagery from ISRO's Bhuvan portal, a remote sensing data visualization platform. These images were processed by the AI model, Segment Anything Model 2, to provide a ground truth.
The results for horizontal mapping were impressive. In planned residential communities across five major Indian cities, GlobalBuildingAtlas was nearly perfect, achieving a completeness coefficient of over 99%. In cities like Bangalore and Calcutta, the digital map matched the actual building. This indicates that GBA is an exceptionally reliable tool for understanding neighborhood density or the scale of urban sprawl.
However, the vertical results told a different story. Researchers discovered a systematic negative bias, meaning GBA almost always underestimated building heights. Errors became more significant as building heights increased. For example, for a 266-meter skyscraper in Mumbai, GBA estimated the height at only 113 meters, representing a discrepancy of over 150 meters. For mid-rise buildings, errors often exceeded 40%.
GBA uses monocular optical imaging, meaning it estimates height from two-dimensional photographs rather than using direct three-dimensional measurements. Although AI can analyze shadows and perspective to estimate height, it often gets confused due to complex roof shapes, nearby trees, or the close proximity of buildings in densely populated Indian cities.
Previous global datasets, such as those released by Microsoft and Google, primarily focused on two-dimensional outlines. GBA represents an ambitious attempt to add the third dimension—height—which is crucial for modern sustainable development goals. However, the ISRO study highlights a serious limitation. The AI models used to create GBA were largely trained on data from Europe and North America. Since architectural styles, building materials, and urban density in the Global South, especially in India, differ greatly, the AI struggles to correctly interpret the scenes.
Furthermore, GBA uses a simplified three-dimensional representation known as Level of Detail 1, which treats each building as a flat block, ignoring complex spires, pitched roofs, and helipads that characterize many Indian landmarks. The study concludes that while GlobalBuildingAtlas is a valuable resource for understanding the horizontal layout of our cities, it should be used with extreme caution when vertical accuracy is required. This work serves as a vital warning to society, as when designing smart cities, planning disaster response, or calculating rooftop solar energy, it is necessary to ensure that the data reflects reality. By identifying these height errors, the ISRO team provided a roadmap for future improvements, suggesting that integrating more laser data from satellites like ICESat-2 could eventually lead to a digital world as tall and complex as our own.
Apple may modify its usual smartphone launch schedule. Information from supplier Pegatron suggests that the iPhone 18 Pro models will be available in the second half of 2026, while the standard version would only be released in the first quarter of 2027.
This calendar change would also affect other devices planned for the next generation. The iPhone 18e and a new model called iPhone Air are considered candidates for later release, as is a possible foldable iPhone from the Ultra line, expected among this autumn's premium products.
The main reason for this shift would be linked to limitations in component supply, specifically memory and chips. With more restricted parts availability, both manufacturers and customers like Apple would be restructuring production peaks, prioritizing the more expensive devices that generate higher profit margins.
The idea of a staggered launch is not new. Last year, analyst Ming-Chi Kuo and the portal The Information had already mentioned that Apple was considering concentrating the launches of its most advanced equipment for the fall of 2026.
More recently, another company supplier indicated that one of its clients was postponing a smartphone launch until early 2027. This new report, published by Economic Daily News and reported by MacRumors, adds an important detail: statements made by Pegatron executives during a presentation on the company's results.
According to the meeting report, the manufacturer confirmed that the iPhone 18 Pro series devices should be launched in the second half of this year. On the other hand, the conventional iPhone 18 would have its launch rescheduled for the first months of 2027.
This division in the calendar is consistent with previous rumors about the iPhone 18e and the iPhone Air. These devices are also expected to be part of the batch arriving after the professional models.
The strategy would be directly related to the difficulties encountered in component supply. According to the Pegatron executives' report, the shortage of memory and chips would be altering corporate behavior during periods of high demand.
In this context, companies like Apple would be optimizing the distribution of available components among various products. Priority would be given to higher-value smartphones, as these have superior margins and consequently represent a more efficient use of a limited stock of parts.
If this schedule is confirmed, Apple would cease concentrating the entire next generation of iPhones in a single annual period. The most sophisticated models would be presented initially, while the more accessible positioning devices would be reserved for a second phase in 2027.
Among the premium launches expected for the second half is also the rumored Ultra foldable iPhone. The inclusion of this device alongside the Pro line would reinforce the possibility of the company dedicating its main showcase of 2026 to higher value-added products.
Google has implemented an extensive restructuring of its artificial intelligence division, relocating parts of the Google DeepMind teams to the corporate structure and concentrating strategic decisions under the leadership of Koray Kavukcuoglu. This information was reported by Reuters on Wednesday, the 12th.
This change occurs amid intense competition, where Google seeks to regain lost ground against competitors such as Anthropic and OpenAI. Furthermore, the move reduces the historical autonomy of the lab, established after its acquisition in 2014, aligning Gemini's development more closely with the company's commercial objectives.
In parallel, Sergey Brin resumed his influence on model training efforts, advocating for greater acceleration to reach technological frontiers. However, Google faces challenges, including delays in the launch of the next major version of Gemini and restrictions on available computational capacity.
The transformation gained public visibility on August 5th, the date Google announced changes in the management of Google DeepMind. Demis Hassabis stepped down from the executive leadership of the lab and assumed the presidency, while his successor, Koray Kavukcuoglu, took on the central role in the unit's direction.
This new setup formalized a power transition that was already underway. Kavukcuoglu had held the position of Google's principal artificial intelligence architect since 2025, a role created specifically to expand Gemini's presence in the company's products. Sources close to the operation informed Reuters that his influence over the model's direction has grown significantly since then.
Kavukcuoglu's position also brought him closer to the divisions responsible for AI revenue generation. Even before moving from London to Mountain View, he was seen as the main link between DeepMind and Google Cloud, a sector considered vital to Google's AI business.
The new arrangement impacted other prominent names. Jeff Dean and Oriol Vinyals, who were part of the original Gemini technical team, left the company to join a new startup. The departure of these executives raised questions about the future of groups that had greater freedom to conduct research outside of direct model training.
During a general meeting held on August 6th, employees were notified that certain non-technical teams would be removed from the DeepMind structure, reporting directly to the corporate organization of Google. This measure reinforced among lab members the feeling that their independence was diminishing.
Management attempted to calm these concerns, assuring that the lab's daily activities would remain essentially unchanged. Hassabis would continue to focus on long-term issues related to the science and social impacts of artificial intelligence, while Kavukcuoglu would maintain control over Gemini.
However, this change comes amidst growing pressure for results. According to reports gathered by Reuters, internal tests forced Google to postpone the arrival of the next major version of Gemini by two months due to lower results compared to those achieved by competitors in areas such as programming.
Recent history contributed to this pressure. Gemini briefly led among competing models after an update released in November 2025, but subsequently lost ground against advances made by rival companies.
Sources involved in development attributed part of this difficulty to structural problems. Limitations in computational capacity availability and disagreements among project leaders hindered the concentration of resources in crucial areas, such as programming. Internal bureaucracy was also cited as a factor contributing to a slower release pace than that of more agile competitors.
In this scenario, Brin informally intervened with the teams responsible for model training. According to someone familiar with his involvement, the co-founder spent months pressuring employees for a faster approach to state-of-the-art systems.
Brin's influence also extended to resource allocation. Among the topics that received his attention was the concept of recursive self-improvement, linked to the possibility of AI systems improving their own capabilities without human intervention.
Although he no longer holds an executive position, Brin maintains influence due to his status as a co-founder. Reuters mentioned that he also adopted a posture of internal mobilization after the launch of ChatGPT in 2022, when OpenAI's progress generated a strong internal reaction at Google.
For his part, Hassabis had reduced his direct involvement in Gemini's management after assisting with the model's launch in 2023. Sources cited by Reuters indicated that development responsibilities were progressively transferred to his subordinates.
The concentration of authority in Kavukcuoglu may also influence the dynamic between DeepMind and Google Cloud. A report obtained by Reuters suggests that cloud computing executives expect the change to minimize conflicts related to the distribution of artificial intelligence chips, known as TPUs.
Thus, the new design positions model strategy, Gemini's integration into products, and commercial requirements within a structure closer to Google headquarters. A company spokesperson confirmed to Reuters that Kavukcuoglu will have the final say on important decisions involving DeepMind.
