Cloud Hosting

You may have just updated your iPhone to iOS 17.1, but Apples already hard at work the next big iOS update. The company is currently beta testing iOS 17.2, which might just offer the biggest changes since...well, iOS 17.0.

Heres whats new:

One headliner this time around is Journal, Apples first-party journaling app. The company announced Journal during its WWDC iOS 17 presentation, but it didnt make the final cut for iOS 17.0 or iOS 17.1.

Apple differentiates Journal from other similar apps by intelligently pulling from things you did on your iPhone that day to create entires. For example, Journal takes photos you snapped, places you visited, music you listened to, and workouts you did, and automatically adds those to generate a journal entry. From there, you can journal away, either with Apples prompts, or with whatever you want to write.

The new update also brings iMessage Contact Key Verification to the iPhone. Apple previously teased this feature too. For those who face extraordinary digital threats (think journalists and politicians who are hacking targets), this feature confirms whether the person theyre iMessaging is really who they say they are. If your iPhone detects an unrecognized device has breached the cloud servers and infiltrated your conversation, youll receive an alert. Theres even a code you can use to confirm the other persons identity.

Those updating to iOS 17.2 will find two key playlist features in Music. Now, you can collaborate with others on playlists, so a group of you running iOS 17.2 can all contribute to the list of songs, including what order they play in. In addition, iOS 17.1's Favorites, which replaces the useless Love function, are now added to a new Favorites playlist. It finally makes revisiting the songs you marked as a favorite easy.

The new update now gives you the option to disable Apple Music listening history for any given Focus. So, if youre listening to ambient music while you have a Work Focus enabled but you dont want it added to your history, none of that playback will be recorded to your account.

Your iPhone 15 Pros Action Button is cool, and can be used in a host of different ways. Now, you can add translation to that list: If you choose the Translate option in iOS 17.2, you can use the Action Button to quickly pull up the Translate app.

This ones another delayed iOS 17 feature: In addition to the usual Tapback reactions you can use, iOS 17.2 lets you react with either an emoji or a sticker. And, since iOS 17 lets you turn just about anything into a sticker, you can react to messages with just about anything.

You now have three new Weather widget options to choose from: Details, Daily Forecast, and Sunrise & Sunset. Theres also now a digital clock widget.

There are also a couple of smaller features Apple added to iOS 17.2. You now have a rainbow text option when making a Contact Poster, and Apple Books has a new Fast Fade feature when turning pages.

Read the original here:
Everything New in iOS 17.2 - Lifehacker

Read More..

Solutions Reviews Contributed Content Series is a collection of contributed articles written by thought leaders in enterprise tech. In this feature, Rockwell Automations Albina Ortiz offers commentary on securing data collection and analysis at the edge.

Despite being around since the 1990s, manufacturers have only recently adopted edge computing, driven by the advance of digital transformation and the expansion of Internet of Things (IoT) device connectivity to enterprise networks. This dispersed computing model brings computation and data storage closer to the networks data source or edge (hence the name) rather than relying solely on centralized cloud servers. Its ability to bring storage and computation closer to machines and locations that need it is now proving its worth to C-level executives.

While edge computing offers numerous advantages, it also raises significant security concerns. Securing data collection and analysis at the edge is crucial in todays connected world, especially as more devices and sensors are deployed in remote or distributed environments and may be more vulnerable to physical attacks or unauthorized access. In this article, we will explore why edge computing is both a blessing and a challenge from a cybersecurity perspective, why securing a distributed network requires careful consideration, and how manufacturers can help protect themselves from breaches.

From a cybersecurity perspective, edge computing has many beneficial attributes for enterprises, including improved response time, bandwidth optimization, and decreased security concerns. Lets highlight some of the main benefits:

By moving security applications closer to the data and devices they are protecting, edge computing reduces latency, improving their performance and making it easier to detect and respond to threats in real time. Edge computing also reduces bandwidth utilization and brings data storage and processing closer to the user, delivering quicker response times than cloud computing and improving performance. By processing data closer to where its generated, applications can respond more quickly, making it ideal for real-time and latency-sensitive applications such as IoT devices, autonomous vehicles, high-quality live video streaming, and augmented reality (AR).

Edge computing helps improve the overall security posture by distributing security controls throughout the network, making it more difficult for attackers to gain a foothold and compromise the entire system. In addition, deploying real-time threat detection and response systems using edge computing can identify and block attacks before they reach central data centers. With data no longer being sent over long distances, edge computing aids in security enhancements, shielding sensitive data from potential cyber threats and reducing the risk of data breaches. Edge computing can also authenticate and authorize devices before they are allowed to connect to the network, helping prevent unauthorized devices from accessing sensitive data and systems.

Computing at the edge can help improve data privacy by reducing the amount of data that needs to be transmitted to central data centers and encrypting the data sent, making it more difficult for threat actors to steal or intercept sensitive data. Enhancing data privacy and security through edge computing keeps sensitive data localized, reducing the need to transmit it over public networks. This is crucial for industries like healthcare and finance, which handle sensitive information.

Edge computing can help reduce an organizations attack surface by offloading some processing and storage from central data centers. This can make it more difficult for attackers to find and exploit vulnerabilities.

Zero-trust security models implemented using edge computing further enhance overall security by verifying all users and devices identities and authorization before granting access to resources. This can help prevent attackers from gaining access to sensitive data and systems, even if they have compromised a users account or device.

The flip side of edge computing is that it introduces several security risks that enterprises must consider and address. These risks stem from edge networks distributed nature, edge devices diversity, and the potentially remote and less secure environments where these devices may be deployed. Lets take a close look at these challenges:

The nature of edge computing involves distributing devices and applications to the edge of the network, which can make them more challenging to secure and manage. The diverse range of devices, from simple sensors to powerful servers, can make it difficult to develop and implement effective and adequate security controls for all devices.

Achieving interoperability and maintaining data consistency across distributed edge nodes, devices, and platforms is also a challenge. Ensuring that all nodes have access to the most up-to-date information without introducing conflicts is complex, so making sure that different devices and protocols can work together seamlessly is an ongoing concern.

Edge devices typically have limited computational power, memory, and storage compared to cloud servers. This constraint can make it challenging to run resource-intensive applications, handle large datasets at the edge, and difficult to deploy robust security measures like intrusion detection systems or complex encryption algorithms.

Additionally, edge devices may not receive timely security patches or updates due to their remote locations or constrained resources, so vulnerabilities may persist longer on these devices, making them attractive targets for attackers. Edge devices, especially personal devices, can often be misconfigured, creating additional potential security vulnerabilities.

Often located in remote or unsecured locations, edge devices can be more vulnerable to physical attacks. Edge devices exposed to extreme weather conditions, power fluctuations, or other environmental factors can impact their security and reliability. The increased number of secured devices and data points that edge computing facilitates raises the chance of security breaches. Moreover, weak authentication and authorization mechanisms can allow unauthorized users or devices to access and manipulate data or control edge devices, so proper identity and access management are critical.

Managing a distributed network of edge devices can be more complex than a centralized cloud infrastructure, creating new attack vectors and additional entry points for attackers. Threat actors can exploit those devices vulnerabilities to gain network access or steal data.

Edge computing can include a higher up-front cost for the initial acquisition of additional hardware and software, in addition to potential training required to manage and maintain a new solution, despite the savings over the longer term.

The businesss specific needs must be considered when securing a distributed network. For example, a company that relies on real-time data processing may need to implement stricter security controls to help protect against data breaches and denial-of-service attacks. Securing a distributed network presents a unique set of challenges and vulnerabilities that differ from those of traditional centralized networks and require careful consideration.

Why is this? In short, complexity, the variety of devices with different operating systems and security vulnerabilities, the demand for compliance, and network visibility.

Distributed networks are complex and have no central point of control or authority. They involve several technologies, vendors, locations, and, often, multiple perimeters and consist of various devices, nodes, or endpoints with different operating systems and inherent security vulnerabilities. Remote devices, such as laptops and mobile devices, can be more challenging to secure than devices on-premises. This extensive attack surface provides attackers with more opportunities to exploit weaknesses, and decentralization can make it harder to enforce security policies, monitor network activity, and respond to threats in a coordinated manner. Securing each component and the channels between them is an essential consideration.

The dynamic nature of distributed networks means they are constantly changing, making it difficult to maintain visibility into all aspects of a distributed network, keep up with security risks, and ensure that security controls are adequate. Enterprises need to be conscious of these dynamics when considering security.

Distributed networks often handle sensitive data, and many industries have regulatory requirements (e.g., GDPR, HIPAA) that mandate strict data protection measures. Achieving and maintaining compliance across a distributed network can be complex and demanding. Meeting these regulatory requirements is a crucial consideration for securing a distribution network.

Securing a distributed network effectively and providing protection from security breaches requires a comprehensive security strategy that identifies specific security risks and implements appropriate controls to mitigate those risks. The security strategy requires regular reviews and updates to remain effective. Below are some key elements that form part of a robust cybersecurity strategy for edge computing networks.

A zero-trust security model verifies the identity and authorization of all users and devices before granting access to resources. This can help to prevent attackers from gaining access to sensitive data and systems, even if they have compromised a users account or device.

Strong authentication and authorization mechanisms verify the identity of edge devices and users accessing them. These controls help prevent unauthorized users and devices from accessing sensitive data and systems and can include certificates, keys, or biometric authentication. Using role-based access control (RBAC) to define and manage permissions for devices and users limits access to only what is necessary for each entity.

Implementing security segmentation to isolate different parts of the network and edge devices from critical systems can limit the damage caused by a successful attack and help prevent attackers from moving laterally through the network if they can compromise one device or application. Installing firewalls and intrusion detection systems (IDS) enables monitoring and controlling traffic between segments.

Encrypting data at rest and in transit can help to protect it from being intercepted or stolen. It is crucial to ensure that data is encrypted and transmitted between edge devices and data storage or processing centers using secure protocols like MQTT, CoAP, and AMQP. Encrypting data stored on edge devices, especially with persistent storage, using encryption algorithms and keys that meet industry standards, helping protect data in place.

Developing a robust patch management process to keep edge devices up to date with the latest security patches and firmware to address known vulnerabilities is essential for managing them as they are discovered and making it more difficult for attackers to exploit them. Hardware-based security features such as Trusted Platform Modules (TPMs) can store encryption keys and help ensure the devices integrity.

Implementing data integrity checks, such as hash functions, can ensure that data hasnt been tampered with during transmission or storage while validating data at the edge filters out potentially malicious or erroneous data before analysis. Defining clear data retention and disposal policies at the edge helps ensure that data is not stored longer than necessary to reduce the risk of unauthorized access.

Implementing real-time monitoring of edge devices and networks helps detect and respond to security incidents quickly and effectively. Using intrusion detection systems and anomaly detection can identify suspicious activity while maintaining detailed logs of device and network activities that can be analyzed for security incidents and breaches. Its essential to have a well-defined incident response plan to address security breaches promptly, including procedures for identifying, containing, mitigating, and recovering from security incidents.

Employee awareness of security best practices and the importance of data security is vital to understanding potential threats. It can help users avoid making mistakes that could lead to security breaches. Training employees and personnel working with edge devices and data is essential to an enterprises security strategy.

Staying informed about relevant data privacy and security regulations, such as GDPR, HIPAA, or industry-specific standards, is another crucial element of a security strategy to ensure compliance with these regulations in edge computing environments.

Edge computing offers enterprises many benefits, including increased data processing and speed, reduced latency, improved data privacy, and bandwidth optimization. It also brings unique security challenges that must be addressed to protect sensitive data.

Securing a distributed network is a complex task due to its expansive and diverse nature and the need to address various technical, operational, and compliance challenges. Securing data collection and analysis at the edge is an ongoing process that requires a proactive approach, especially as more devices and sensors are deployed in remote or distributed environments.

To mitigate security risks in edge computing environments, organizations should adopt a comprehensive security strategy that includes strong authentication, encryption, intrusion detection, regular patching, secure device provisioning, and continuous monitoring. Additionally, security best practices should be integrated into the design and deployment of edge solutions to minimize vulnerabilities and safeguard critical assets and data. Collaboration with experts in cybersecurity and regular security audits can also help ensure the effectiveness of security measures.

Related

More:
Securing Data Collection and Analysis at the Edge - Solutions Review

Read More..

Edge AI and machine learning have become increasingly important in industrial technology settings. With the rise of IoT sensors and smart devices, there is a growing need for edge AI solutions to process data locally and make real-time decisions without relying on cloud services.

TDKs Qeexo AutoML platform was designed to help developers and companies easily implement AI solutions at the edge without the need for extensive machine learning expertise. The goal is to provide a user-friendly interface and automated feature engineering. According to TDK, Qeexo AutoML can significantly reduce the time and resources needed to develop and deploy edge AI applications.

The need for privacy and security is also important for data processing in industrial applications. With edge AI, data is processed locally without being transmitted to a remote server, ensuring better protection for sensitive data.

Edge AI can improve efficiency and reduce latency in various industries such as manufacturing, healthcare, and transportation. By bringing intelligence closer to the source of data, edge AI enables faster decision-making and real-time monitoring. Altogether, edge AI solutions can help reduce costs and improve performance.

We caught up with Sang Lee, CEO of TDK Qeexo, to dig deeper into the issues and advantages of using edge AI.

Sang Lee: The adoption of Edge AI faces two primary challenges: first, it creates high-performance models that can run efficiently on edge hardware, and second, it can address the mass deployment challenges associated with customizing and localizing models for optimal performance in diverse environments.

Sang Lee: Artificial intelligence is vital for IoT and smart devices because it transforms data into actionable insights, enabling real-time decision-making, predictive maintenance, enhanced security, energy efficiency, and personalized user experiences. AI's automation and scalability benefits make IoT devices smarter, more efficient, and adaptable, driving cost savings and reducing environmental impact. It's the cornerstone for realizing the full potential of the IoT ecosystem, revolutionizing industries, and improving the way we interact with our connected environments.

Sang Lee: Edge AI refers to the deployment of AI algorithms and models directly on edge devices, such as IoT devices, rather than relying on a centralized cloud server for processing. It helps in making real-time decisions by bringing AI capabilities closer to the data source, where data is generated, and decisions need to be made.

Sang Lee: Qeexo AutoML streamlines machine learning workflows by automating the most labor-intensive tasks, including data cleaning, sensor and feature selection, model choice, hyperparameter optimization, model validation, conversion, and deployment. This automated process enhances efficiency and saves valuable time, as it evaluates numerous options behind the scenes, identifying the most suitable models for your data.

Sang Lee: Processing AI on the edge device allows for immediate handling of all raw sensor data at the source, eliminating the need to transmit raw data to a remote server for processing. This results in enhanced power efficiency and reduced latency.

Sang Lee: Edge AI can process sensitive data locally, without sending it to external servers. This enhances data privacy and security, which is crucial for applications like surveillance and healthcare.

The rest is here:
Industry Voices: Making Data Actionable with Edge AI - Design News

Read More..

On October 17, 2023, the US Department of Commerces Bureau of Industry and Security (BIS) released two interim final rules (collectively, the October 2023 IFRs available here and here) amending the Export Administration Regulations (the EAR) to further strengthen export controls on advanced computing items, semiconductor manufacturing equipment, and items that can support end uses related to the development and production of supercomputers, advanced-node integrated circuits and semiconductor manufacturing equipment. The long-awaited October 2023 IFRs come more than one year after BIS issued rules on October 7, 2022 (the October 2022 IFR) targeting Chinas advanced computing and semiconductor sectors and update, expand and clarify those rules in several key ways. Our previous blog post on the October 2022 IFR can be found here. BIS also issued a final rule adding two Chinese entities and their affiliates (13 Chinese entities in total) involved in the development of AI-capable advanced computing chips to the Entity List with a footnote 4 designation (the Entity List FR).

According to the BIS press release and the preambles to the October 2023 IFRs and the Entity List FR, the latest changes are intended to address the national security and foreign policy concerns in countering Chinas military-civil fusion strategy and the development of advanced or frontier AI capabilities that can lead to improved weapons of mass destruction (WMD) and advanced conventional weaponry while minimizing unintended impact on trade flows.

The updates and clarifications in the October 2023 IFRs are extensive and also include BIS responses to the many comments on the October 2022 IFR that provide noteworthy guidance and interpretation on several topics, including quite expansive jurisdictional interpretations with respect to semiconductor supply chains in the context of the end-use controls of EAR 744.23 and the scope of the Entity List Foreign-Direct Product (FDP) rules. A detailed review of all the comments is beyond the scope of this post, but should be a priority for all US and non-US participants in the semiconductor and supercomputer industries.

The October 2023 IFRs take effect on November 17, 2023 (except for a temporary general license which is effective on October 25, 2023). Note that BIS is seeking public comments on certain topics that have not been fully addressed by the October 2023 IFRs, including with respect to access to cloud/infrastructure-as-a-service (IaaS) to develop AI foundation models and deemed exports/reexports. The comment period ends on December 18, 2023.

I. Background

The earlier October 2022 IFR focused on: (i) item-based and end-use-based controls for certain advanced computing integrated circuits (ICs), commodities incorporating such advanced ICs, and supercomputers; and (ii) item-based controls for certain semiconductor manufacturing equipment (SME) and end-use-based controls related to the development and production of SME. The October 2023 IFRs expand and clarify these through an Advanced Computing/Supercomputing IFR (the AC/S IFR) and a Semiconductor Manufacturing Equipment IFR (the SME IFR).

II. Key Changes in the October 2023 IFRs

The AC/S IFR creates a new structure and updates the technical parameters used for identifying items controlled under ECCN 3A090 to ensure ICs for AI training are controlled. Structurally, the revised 3A090 replaces former paragraphs .a.1 through a.4 with a simplified paragraph .a (for the most powerful data center ICs) and a new paragraph .b (covering additional, less powerful but still advanced ICs that could be used to train large-scale AI systems). Substantively, the revised entry removes interconnect bandwidth as a technical parameter and adds performance density as a new parameter. These changes are intended to counter workarounds to the former controls involving the use of a larger number of smaller data center AI chips containing the same power as one restricted chip.

Most importantly, the AC/S IFR (i) excludes certain non-datacenter ICs from control (e.g., certain high end gaming chips) and (ii) replaces bits x TOPS with Total processing performance (TPP) values, defining clear, objective criteria that can be used to calculate the TPP value. The AC/S IFR further refines the scope of control through the addition of new notes to 3A090 and revisions of the technical notes.

The October 2022 IFR used the catch-all criteria or identified elsewhere on the CCL that meet or exceed the performance parameters of ECCNs 3A090 or 4A090 in describing the scope of certain controls (e.g., 742.6(a)(6)). BIS received comments raising concerns that this type of language is confusing, creates potentially dual ECCNs, and makes compliance burdensome.

To address these concerns, the AC/S IFR creates new .z paragraphs for nine ECCNs to affirmatively capture items that BIS has determined have performance characteristics or functions that meet or exceed the relevant performance parameters. The nine ECCNs are 3A001.z, 4A003.z, 4A004.z, 4A005.z, 5A002.z, 5A004.z, 5A992.z, 5D002.z, and 5D992.z.

The AC/S IFR also requires exporters to identify .z items at the items level classification in the electronic export information filing in the Automated Export System and, in certain cases, on the commercial invoice for export clearance.

The SME IFR removes ECCN 3B090. Controls on SME formerly under ECCN 3B090 are expanded and placed under ECCNs 3B001 and 3B002 (with a view to these being controlled multilaterally in the future). A number of additional types of SME and equipment for manufacturing SME are now controlled, including: certain equipment designed for silicon (Si), carbon doped silicon, silicon germanium (SiGe), or carbon doped SiGE epitaxial growth; certain equipment designed for coating, depositing, baking or developing photoresist formulated for EUV lithography; semiconductor wafer fabrication cleaning and removal equipment; and inspection equipment designed for EUV mask blanks or EUV patterned masks, amongst others. These revisions to ECCNs 3B001 and 3B002 also prompted conforming changes to ECCNs 3D001, 3D002, and 3E001. For software and technology related to newly controlled SME described under 3B001, Regional Security (RS) and National Security (NS) controls have been added. The newly controlled equipment and related software/technology have been determined by BIS only to be used (with limited exceptions) for fabricating logic ICs with non-planar transistor architecture or with a production technology node of 16/14 nanometers or less. Companies should carefully review the new and revised ECCNs.

The SME IFR states that BIS licenses for equipment classified under ECCN 3B090 remain valid until expiration, unless suspended or revoked. At the same time, exporters are required to list the new ECCNs on any export clearance documentation filed after November 17, 2023.

The SME IFR provides that there will be no de minimis level for certain foreign-made lithography equipment (and specially designed parts) that are now classified under ECCN 3B001.f.1.b.2.b when destined for use in the development or production of advanced-node integrated circuits (Advanced-node ICs), except if the country from which the foreign-made item was originally exported also controls the item (currently only Japan). This means that otherwise any amount of US origin content will make a foreign-made lithography equipment subject to the EAR under this ECCN.

The AC/S IFR broadens the country scope with respect to the advanced computing FDP rule to Country Group D:1, D:4, and D:5 countries that are not also specified in Country Groups A:5 or A:6 (currently, countries that fall within both Country Groups D:1, D:4, and D:5 as well as Country Groups A:5 and A:6 include only Cyprus and Israel). The country group for this rule is further expanded to be worldwide when the foreign-made direct product is for an entity headquartered in, or whose ultimate parent entity is headquartered in, either Macau or a D:5 country.

The AC/S IFR clarifies that the model certificate published in the October 22 IFR and which appears in supplement no. 1 to part 734 of the EAR may be used for all FDP rules. The AC/S IFR clarifies that the model certificate may be provided by anyone in a supply chain and flows both forward and backward in a supply chain (e.g., from the exporter, reexporter, or transferor to another entity in the supply chain or from a consignee back to an exporter, reexporter, or transferor).

The October 2023 IFRs expand the list of countries for which licenses are required for National Security (NS) and Regional Stability (RS) reasons. This is apparently intended to close a loophole in exports to offshore affiliates of Chinese/Macau companies.

Specifically, under the revised controls in 742.4 and 742.6, a license requirement now extends to:

Applications that involve transactions falling under categories (i) and (ii) above will be reviewed consistent with license review policies in EAR 744.23(d), except applications will be reviewed on a case-by-case basis if no license would be required under part 744 of the EAR rule. For applications falling under category (iii) above, if the items are for Macau or a D:5 country, then the applicable license review policy is a presumption of denial; if the items are for non-D:5 countries, the applicable license review policy is generally a presumption of approval, unless to an entity headquartered in, or whose ultimate parent entity is headquartered in, either Macau or a D:5 country, in which case the applications will be reviewed under a presumption of denial policy.

Consistent with the October 2022 IFR, the license requirements above do not apply to deemed exports/reexports (although BIS seeks comments on this.)

The AC/S IFR expands the scope of the end-use controls in EAR 744.23 by adding two new advanced computing end-use controls under new EAR 744.23(a)(3) that impose a license requirement on:

The AC/S IFR also expanded the country/destination scope of the supercomputer end-use controls and Advanced-node IC end-use controls to Macau and any D:5 country (not limited to China).

The SME IFR further reorganizes and simplifies the end-use controls adopted in the October 2022 IFR. The changes are shown in the below table included in the SME IFR.

We note the following key takeaways from the changes to 744.23 :

The October 2023 IFRs reformat, revise and clarify the US Person restrictions under EAR 744.6(c), incorporating guidance provided by BIS under FAQs issued since the October 2022 IFR, as follows:

In short, US persons are prohibited from knowingly shipping, transmitting, or transferring (in-country), facilitating these activities, or providing services in connection with the following items, end-uses, or end-users:

No license exceptions apply to these prohibitions. The AC/S IFR creates new sections to supplement no. 2 to part 748 of the EAR with guidance on how to apply for a license under the US person restrictions.

License applications will be reviewed with a presumption of denial for Macau and D:5 countries, except activities involving a foreign-made item not subject to the EAR that performs the same function as an item subject to the EAR which will be reviewed with a presumption of approval. All other applications will be reviewed with a license review policy of case-by-case.

The October 2023 IFRs create two new TGLs as follows:

This AC/S IFR TGL overcomes the new RS controls in EAR 742.6(a)(6)(iii) to authorize the export, reexport, and transfer of certain otherwise controlled items to or within D:1, D:4, or D:5 countries (excluding countries also specified in Country Group A:5 or A:6) where the recipient is located in Macau or a D:5 country but is not headquartered in and whose ultimate parent entity is not headquartered in Macau or a D:5 country to continue or engage in integration, assembly, inspection, testing, quality assurance, and distribution of items covered by the product scope of the TGL, provided for the ultimate end use outside of a D:1, D:4, or D:5 country (excluding countries also specified in Country Group A:5 or A:6) by entities that are not headquartered in or whose ultimate parent company is not headquartered in Macau or a D:5 country.

The AC/S IFR TGL does not overcome the license requirements triggered by the involvement of a party on the Entity List or the Military End-User List or knowledge of any other prohibited end-use or end-user (with some exceptions under 744.23).

The AC/S IFR TGL is valid for two years and expires on December 31, 2025.

This SME TGL authorizes the export, reexport, or transfer (in-country) of (i) non-EAR99 items subject to the EAR and controlled only for AT reasons, (ii) to manufacturing facilities in a D:5 country or Macau, (iii) for the development or production of parts, components, or equipment of certain Category 3B ECCNs specified in EAR 744.23(a)(4), provided (iv) at the direction of companies headquartered in the United States or an A:5 or A:6 country and not majority-owned by an entity headquartered in, or whose ultimate parent entity is headquartered in, either Macau or a D:5 country.

According to the SME IFR, this TGL is intended to give SME manufacturers in the United States and A:5 and A:6 countries additional time to identify alternative sources of supply outside of restricted countries or to obtain BIS licenses to continue manufacturing front-end IC production equipment.

The SME TGL cannot be used for the indigenous development or production of Category 3B tools or parts in a D:5 country or Macau at the direction of an entity headquartered in, or whose ultimate parent entity is headquartered in, either Macau or a D:5 country. This TGL also does not overcome the license requirements triggered by the involvement of a party on the Entity List or the Military End-User List or knowledge of any other prohibited end-use or end-user.

The SME TGL is effective immediately, valid for two years and expires on December 31, 2025.

The AC/S IFR creates a new License Exception NAC (at EAR 740.8) authorizing exports, reexports, and transfers of (1) certain less powerful but still advanced ICs under the new ECCN 3A090.b and (2) certain 3A090.a commodities that are not for data centers but do have a TPP of 4800 or more. License Exception NAC cannot be used for ECCN 3A090.a data center ICs and, with limited exceptions, is generally not available for transactions that are subject to a license requirement under Parts 744 or 746 of the EAR or for transactions involving a military end-use/user.

For exports and reexports (excluding in-country transfers) to Macau or a D:5 (arms embargoed) country, prior notification of 25 calendar days must be provided to BIS via STELA prior to the first export or reexport. Notification requirements do not apply for transfers (in-country) within D:1 or D:4 countries. This notification requirement is intended to enable the US Government to collect information about the export of less advanced ICs. BIS will publish a further notice when STELA is ready to accept License Exception NAC notifications.

The October 2023 IFRs make commodities classified under ECCNs 3A991.p and 4A991.l eligible for License Exception Consumer Communications Devices (CCD), add eligibility for License Exception TMP for temporary exports of advanced computing items under 3A090 and 4A090 for inspection, test, calibration and repair, make other adjustments to various license exceptions related to the new .z ECCN paragraphs, amongst other changes.

The SME IFR introduces two newly defined terms: advanced-node ICs and extreme ultraviolet.

These two terms are used in the US person restrictions in EAR 744.6, the end-use controls in EAR 744.23, and various ECCNs.

In response to comments to the October 2022 IFR, the AC/S IFR adds five new red flags to the Know Your Customer Guidance in supplement no. 3 to part 732 of the EAR to help the public ensure compliance with the new rules. These include a red flag for knowledge of future intent to develop or produce restricted supercomputers or advanced ICs. Additional due diligence may be required to address and resolve these red flags before proceeding with a transaction.

III. Entity List FR

BIS added two Chinese entities and their affiliates (13 Chinese entities in total) to the Entity List with a footnote 4 designation for their involvement in the development of large AI models and AI chips for defense purposes.

The Entity List FR includes a savings clause that exempts shipments from the new license requirements associated with the Entity List party that were en route aboard a carrier to a port of export, reexport, or transfer (in-country) on October 17, 2023, pursuant to actual orders.

***

Despite some simplifications, the new restrictions remain extensive, highly complex and technical. Companies in the advanced computing, semiconductor manufacturing equipment and supercomputer sectors should carefully review the new October IFRs to assess the impact on their operations. We will continue to monitor developments and are available to answer questions.

Read more:
BIS Strengthens Export Controls on Advanced Computing Items ... - Sanctions & Export Controls Update -

Read More..

TL;DR

Google and Qualcomm have partnered to bring powerful generative AI experiences to Android phones. The two companies announced their tie-up to bring more on-device AI to Android devices during the Snapdragon 8 Gen 3 launch. Googles state-of-the-art foundation AI models will run on Qualcomms Hexagon NPU. These are likely the same foundation models that also run on the Pixel 8 Pro and power AI features like on-device recorder summaries, Smart Reply in Gboard, and the upcoming Video Boost feature.

With Googles AI models running on-device on the Snapdragon 8 Gen 3, more Android phones will be able to perform on-device AI tasks, making way for new features, performance improvements, and power savings.

Among the first generative AI models you see running on Snapdragon will come from none other than Google. Over the past few months, researchers at Google have been working to take their massive next-generation large language models and distill them to fit on a mobile device. Soon, youll be able to do more on-device with Google applications than ever before, said Alex Katouzian, SVP and GM of Mobile, Compute, and XR at Qualcomm.

Googles VP of engineering for Android, Dave Burke, also joined the conversation on stage during the Snapdragon 8 Gen 3 launch to confirm that the companys partnership with Qualcomm will enable complex AI models to run on-device on upcoming Android flagship phones, removing the need for internet connectivity and round trips to cloud servers.

Simply put, we can now expect smarter and faster AI processing on premium Android phones with the latest Snapdragon chip. The Qualcomm Snapdragon 8 Gen 3 processor supports large language models with over 10 billion parameters running at almost 15 tokens per second. You can read more about the Snapdragon 8 Gen 3 here. Qualcomm also gave us a glimpse at what to expect from its next flagship processor. You can read about the Snapdragon 8 Gen 4 here.

Link:
Google's Gen-AI models are coming to more Android phones with ... - Android Authority

Read More..