Google announces the definitive end of Google Assistant on September 4th, migrating to Gemini

The government is working to restore the functionality of the Indian Regional Navigation Satellite System, known as NaVIC, planning the launch of four satellites in the near future.

Space industry minister Jitendra Singh stated in a written response to the Lok Sabha on Wednesday that satellite NVS-03 is ready for launch after implementing all corrective measures arising from the NVS-02 anomaly. This will be followed by the launches of NVS-04 and NVS-05, and then another satellite is planned to complete the basic level constellation of NaVIC, which consists of seven satellites.

However, he did not specify the timeline for these launches. Jitendra Singh noted that after the failure of the replacement satellite NVS-02 in 2025, the necessary corrective measures were successfully integrated and tested on satellite CMS-03. These measures will also be applied to subsequent satellites, including the series of NaVIC satellites equipped with two-component propulsion systems.

Currently, navigation services can only be provided by IRNSS 1-B, IRNSS 1-L, and NVS-01. Nevertheless, the NaVIC system requires at least four satellites with operational atomic clocks to function effectively, meaning NaVIC no longer provides autonomous positioning services.

In March of this year, the last atomic clock aboard the Indian Regional Navigation Satellite System-1F (IRNSS-1F) stopped working, leading to the paralysis of NaVIC. The first generation of satellites for NaVIC was launched between 2013 and 2018, but the constellation soon began facing problems due to repeated failures of atomic clocks on several IRNSS series satellites.

The NVS series represents the second generation of satellites for NaVIC. Although ISRO successfully placed NVS-01 into its intended orbit in 2023, the launch of NVS-02 in 2025 failed. This delayed the launch of NVS-03, NVS-04, and NVS-05.

When Ernest Mwebaze began developing an artificial intelligence tool for working with various Ugandan languages, he tested both American and Chinese models. The result was unexpected: the Chinese technology proved superior for this project.

According to The New York Times, such cases illustrate how African developers turn to Chinese systems due to lower costs, customization capabilities, and open access.

Mwebaze, who previously worked as a researcher at Google, created a system called Sunflower. This system is used by farmers in Uganda to receive information about weather and cultivation methods in local languages. He stated that the Alibaba model demonstrated higher performance compared to alternatives from Meta and Google for this task.

Mwebaze said: 'We want to create things as cheaply as possible, but so that they work very well.'

This trend is not only observed in Uganda. In Kenya, entrepreneurs are using Chinese models to improve legal and business services. In Nigeria, developers are creating educational tools, and new chatbots adapted to local needs are emerging in Ghana.

One of the key advantages of these tools is the open source code. Unlike closed models, which require payment and adhere to the rules of the creating companies, these systems can be downloaded and modified by the users themselves.

Among the reasons for this preference: for many African developers, the deciding factor is not using the most advanced model on the market, but finding a tool capable of solving the necessary problem. As entrepreneur Moses Kemibaro noted in The New York Times: 'Why use an expensive Ferrari to take children to school when a Toyota hatchback can do the same thing?'

China's presence in the African tech sector began even before the age of artificial intelligence. Chinese companies were already involved in expanding telecommunication networks on the continent, laying the foundation for offering new solutions.

More details:

Nevertheless, African companies continue to use American tools in certain scenarios. Systems from OpenAI and Anthropic remain in demand for technical tasks requiring high precision.

The choice is also influenced by political aspects. After issues with the Sunflower chatbot's responses regarding China, Mwebaze began looking for alternatives to avoid excessive dependence on a single supplier. He emphasized: 'We must ensure that we are not on the wrong side of geopolitics.'

The African experience shows that the race for artificial intelligence depends not only on who creates the most perfect models. Increasingly, factors such as cost, freedom of use, and adaptability influence the choice of technologies.

Researchers from Unit 42, a division of Palo Alto Networks, have identified three new methods that could allow malware to bypass the protection of passkeys stored in the Google Password Manager in Chrome for Windows.

These attacks, named Pass-ta-key, Silver Pass-ta-key, and Golden Pass-ta-key, do not violate the cryptography of the technology itself, but they exploit vulnerabilities in how the browser manages user authentication and credentials.

According to the researchers, all scenarios depend on one crucial condition: the victim's computer must already be infected with malware. Nevertheless, two of these methods can grant attackers persistent access to the victim's accounts, even after the victim loses access to the original device.

As of early August, the research did not reveal active exploitation examples or registered vulnerabilities in the National Vulnerability Database (NVD).

The first method, called Pass-ta-key, allows malware to use TPM-protected cryptographic identification to request valid authentication without needing to enter a PIN, biometrics, or any user interaction. However, this method does not activate the User Verified (UV) indicator, which notifies the online service of personal authentication confirmation by the user. The effectiveness of this attack, according to the researchers, depends on the implementation of each specific platform.

During testing, GitHub correctly rejected the authentication attempts, requiring verification of this indicator. Meanwhile, eBay accepted these accesses before receiving information from Unit 42 and subsequently fixed the issue.

The second technique, Silver Pass-ta-key, utilizes the device registration process in the Google Password Manager. According to the research, malware forces Chrome to re-register the computer and uses an intermediate stage of this process to install a verification key controlled by the attacker. If the online service accepts this new key as valid, the attacker can continue accessing the victim's account from another computer, even if the original device becomes unavailable. Analysis of the Chromium open source code confirms the existence of this temporary state during the registration process, which can be exploited before hardware verification is complete.

The technique considered the most dangerous is named Golden Pass-ta-key. When used, malware attempts to extract the so-called Security Domain Secret (SDS) from Chrome's memory—a 32-byte cryptographic secret used to decrypt all synchronized passkeys in the Google account. According to Unit 42, obtaining this secret allows for the recovery of both existing and future passkeys created in the same account.

Researchers explain that Google previously eliminated the leakage of this secret in internal logs related to the FIDO protocol. However, the information is temporarily loaded into the browser's memory during device registration and recovery, allowing malware already running on the computer to capture it.

Another aspect raised in the study is that the public documentation for the Google Password Manager allows changing the PIN or deleting all saved passkeys but does not contain information on the mechanism for revoking or replacing the Security Domain Secret in case of compromise.

According to Unit 42, some measures were taken following responsible disclosure of the findings. eBay began correctly verifying the User Verified indicator, preventing the first attack. Google removed the exposure of the Security Domain Secret from Chrome's internal logs.

Despite this, there was no public confirmation before the start of the current month that all methods described by the researchers had been fully mitigated, nor was there information on specific affected Chrome versions.

Despite the discovered issues, Unit 42 asserts that passkeys remain significantly more secure than traditional passwords against phishing attacks and credential reuse. The researchers emphasize that the problem lies not in the cryptography of the technology itself, but in the infrastructure responsible for device management and credential synchronization when the computer is already compromised by malware.

As protective measures, the team recommends that online services mandate the verification of the User Verified indicator, strengthen hardware verification during device registration, reinforce account recovery mechanisms, and reduce the exposure of sensitive information in browser memory and internal logs.