security

a16z Charts compiled recent data indicating that since January 2025, tech recruitment has shown a clear preference for years of experience, while the requirements for the number of skills have decreased, with a change of about 5% to 10%.The charts show that companies are lowering the weight of skill lists when screening positions, and are instead placing more emphasis on candidates' accumulated experience. This change is related to the skill gaps brought about by AI, which is rewriting the safety and employment landscape.

According to the official X account of ZKsync (@zksync), ZKsync announced a security upgrade for its EraVM chain. The core measures include: introducing an instant upgrade framework, extending the execution delay from 3 hours to 24 hours, and developing a second prover, EraBender, to guard against AI-driven vulnerability attacks.In the next 6 months, all Boojum/EraVM chains will gradually phase out the old execution environment, and each chain will formulate and announce its own transition timetable. Users holding funds directly in EOA do not need to take any action; users holding funds through multi-signature, smart accounts, DEX, lending protocols, and other smart contracts need to take action as required after the transition plans of each chain are determined. The ZKsync Atlas chain is not affected by this upgrade.

According to CoinDesk, OpenAI stated that it will provide access to $1 billion worth of Daybreak cybersecurity model subsidies over the next six months, along with training and technical support to help organizations defend against AI-driven cyberattacks, including potential "zero-day" exploits.Daybreak is divided into two tiers, Blue and Red, which are used for routine defense and more sensitive cybersecurity tasks, respectively. Currently, about 2,000 organizations are using it. OpenAI previously disclosed that its new model Astra can discover unknown software vulnerabilities and generate attack plans with less human intervention, raising concerns in the crypto industry about the security capabilities of open-source blockchain software.

According to Securityweek, artificial intelligence security company HiddenLayer has completed a $100 million Series B funding round, led by Delta-v Capital, with participation from Booz Allen Ventures, Microsoft's M12, Morgan Stanley, and Ten Eleven Ventures. To date, the company has raised a total of over $155 million.HiddenLayer focuses on protecting the entire lifecycle security of generative AI, predictive AI, and AI Agents. Its enterprise platform includes features such as AI asset discovery, AI supply chain security, attack simulation, and runtime protection. The new funding will primarily be used to expand the runtime security capabilities of AI Agents, particularly targeting AI programming Agents.

According to Techcrunch, cybersecurity giant Palo Alto Networks announced the acquisition of AI Agent startup Console for $500 million in cash and stock. Founded in 2024, Console uses AI Agent to automate tasks such as password resets, application permission grants, and daily IT troubleshooting, with clients including Ramp, Flock Safety, and Scale AI.Palo Alto Networks plans to integrate Console into its AI-driven security platform Cortex, enabling security teams to investigate and handle security alerts through natural language, achieving more automated enterprise security operations.It is reported that this is also the seventh acquisition completed by Palo Alto Networks this year, having previously acquired Chronosphere at a valuation of $3.35 billion and the cybersecurity startup Koi for $400 million.

Term Labs released a technical report on the security incident, confirming that all fixed-rate loan positions in the affected vault have been restored as of August 25. The incident was limited to the liquid balance within the Term vault, and Term V1 and V2 contracts were not compromised, with the direct lending market remaining open.Attack process: From August 17 to 18, the attacker funded an operational wallet through Tornado Cash and submitted a malicious proposal disguised as a governance proposal, setting the governance delay parameter to zero and removing the window for LP blocking changes. On August 23, the attacker executed two batches of attacks, extracting assets from the ETH and USDC strategies and transferring them to the attacker's wallet. Term Labs stated that it has cooperated with law enforcement, the affected vault and strategies have been closed, and the contracts have been upgraded and migrated before expiration.

In a blog post released on Monday, Anthropic acknowledged that its Claude model had unauthorized access to real computer systems during a cybersecurity assessment, an incident reflecting operational security failures as well as alignment failures in motivation reasoning and intent to harm. Anthropic disclosed in July that the Claude model had breached the systems of three companies because the third-party assessment environment was connected to the public internet, while the model was informed it was in a simulated environment without internet access.Anthropic stated that Claude may have interpreted evidence of real internet access as still being in a simulated environment and was willing to take harmful actions on the real internet to complete the cybersecurity assessment task. Additionally, during tests at the UK AI Safety Institute, after assessors deliberately granted Claude Mythos internet access, the model took unauthorized actions on the live network. Anthropic emphasized that the models involved did not have the cybersecurity protections included in the officially released products.Following the incident on July 30, Anthropic has suspended cybersecurity assessments of pre-release models and introduced stricter protections: tests must run in verified offline sandboxes equipped with real-time monitoring; a new classifier can intercept suspected boundary violations, terminate tests, and notify humans. Anthropic has also expanded the scope of offline monitoring used by internal frontier agents. Previously, OpenAI models had also breached Hugging Face in July to obtain answers for cybersecurity tests, with investigations revealing that about 1,200 agents acted collaboratively through unauthorized message boards.

According to Bloomberg, AI and cloud security startup Upwind Security has completed a new round of financing of $300 million, led by Bessemer Venture Partners, with participation from some existing investors, bringing the post-financing valuation to $3.8 billion.Upwind Security, headquartered in San Francisco, primarily provides cybersecurity services for AI applications and cloud applications. The new funds are intended to expand the AI security capabilities of its core cloud security platform, aiming to be closer to developers.

According to the Securities Times, at the Fourth Cybersecurity (Tianjin) Forum, Zhou Hongyi, the founder of 360 Group, gave a keynote speech, stating that AI has changed the landscape of cyber offense and defense, and traditional solutions are difficult to sustain. He emphasized the need to create a new automated offense and defense system centered on "model governance," using AI capabilities to address AI risks, ensuring the safety of large models, the reliability of generated content, and the credibility of output results.Zhou Hongyi also proposed that there should be lifecycle management for intelligent agents, establishing an auditable, intervenable, and interruptible control mechanism, and called for the cybersecurity industry to quickly form a system for AI vulnerability discovery, automated defense, and intelligent agent security.

OpenAI stated that its upcoming Astra model can autonomously discover previously unknown software vulnerabilities and convert them into usable attack vectors without human intervention, making it the company's first model to reach the "Critical" cybersecurity capability level threshold. In a blog post released on Tuesday, OpenAI mentioned that according to its Preparedness Framework, reaching this level means the model can discover zero-day vulnerabilities and develop usable exploit code in hardened real systems without human involvement, or design and execute attacks based solely on a high-level objective.In testing, Astra achieved a 100% score in benchmark tests for developing exploit code based on known vulnerabilities and discovered two previously unknown vulnerabilities in another internal test. Additionally, the model successfully broke through a hardened browser sandbox and executed commands on the host machine, while gaining root access by exploiting multiple weaknesses in the operating system. OpenAI stated that it has delayed some of Astra's development progress to enhance security measures and plans to make its advanced cybersecurity capabilities available only to selected testers.This capability is particularly relevant to the cryptocurrency industry, as software vulnerabilities can be converted into financial losses within minutes. CoinDesk reported in June that increasingly powerful AI models can compress the process of searching code, discovering misconfigurations, and assembling attacks from days or weeks to machine speed. Security researchers noted at the time that the significant change was not the emergence of new categories of attacks, but rather the dramatically increased speed at which existing vulnerabilities are discovered and exploited.

Core DAO is coordinating an emergency hard fork due to some validators receiving CORE rewards that exceeded the protocol's expected issuance. Core stated in an update that the situation has been brought under control, and "malicious validators" can no longer claim excess rewards. This fork is a forward upgrade and will not roll back the network or revoke any confirmed transactions.Previously, Core indicated in a status update on Monday that the rewards accumulated by a small number of validators were significantly higher than the protocol's expected issuance. The incident was limited to the reward distribution phase, user assets remain secure, and they promised to release a technical review report. Following the incident, several exchanges restricted CORE transfers: Coinbase suspended deposits and withdrawals on the Core network, Bithumb and Coinone suspended deposits and withdrawals, Bitget suspended deposits and withdrawals citing wallet maintenance, and LBank suspended deposits at the request of the project party.Core has not disclosed the amount of excess issued CORE, the duration of the activity, or whether additional tokens have entered circulation, nor have they provided details on the vulnerability that allowed validators to receive rewards. Cointelegraph reached out to Core for further information but had not received a response by the time of publication.

AI security startup AIR has completed a $50 million financing round, led by Sequoia Capital with an initial $10 million in the first round, followed by Greenoaks leading a second round of $40 million, with participation from Swish, Netz, and several angel investors.The company was founded by Yair Saban and Niv Hoffman, senior members of Israel's Unit 8200, and offers an AI agent security platform that can discover AI agents operating within enterprises, continuously review the skills and tools they use, and prevent them from interacting with non-compliant software or external sources, while providing a vetted AI agent plugin and skills marketplace.

According to security company Huntress, hackers are spreading theft malware to cryptocurrency users by forging Google Docs files, hosting malicious files on GitHub, and impersonating Claude.ai pages. Attackers disguise themselves as senior employees of CoinDesk on the social platform X, luring victims to open Google Docs documents containing malicious code under the pretext of inviting them to an online meeting, thereby guiding users to install the malicious program themselves.Mac users face the threat of Atomic macOS Stealer (AMOS), which can steal browser passwords, cryptocurrency wallet data, and Telegram files; Windows users are pushed fake Google API Connector updates, which, once installed, implant NetSupport RAT and a counterfeit Ledger hardware wallet application. Additionally, hackers are placing false advertisements on search engines like Bing, enticing users to visit counterfeit Claude.ai pages and execute malicious commands. The related malware MacSync and SectopRAT can steal cookies, saved passwords, recovery phrases, and payment card information.Security company Socket also discovered 16 malicious extensions targeting Chrome and Edge, capable of emptying EVM, Solana, and Tron wallet assets.

According to the Ontology blog, its core development team discovered potential security vulnerabilities during routine security checks. Due to a high emphasis on the security and integrity of the mainnet, Ontology has immediately suspended block production on the Ontology mainnet to allow the technical team and validators to conduct a comprehensive security review of the network and related components. Ontology stated that no security incidents have been confirmed at this time, and there are no indications that user assets have been lost or stolen. ONT, ONG, and other on-chain assets are currently assessed to be unaffected; this suspension is a preventive security measure rather than a response to confirmed asset losses or ongoing attacks.During the review period, block production will remain suspended, and on-chain transactions cannot be processed. Users do not need to transfer or take any action regarding their ONT, ONG, and other on-chain assets, but should avoid conducting time-sensitive on-chain transactions until the network officially confirms its restoration. The duration of the suspension has not yet been determined, and the team will prioritize the comprehensiveness of the security review over speed; block production will not resume until the network has been thoroughly assessed and confirmed to be secure. The Ontology team is collaborating with validators and ecosystem partners to address this matter and will update progress through official channels, with a separate announcement to be made before and after the network restoration.

Polygon Labs disclosed that it has fixed a batch of security vulnerabilities in its proof-of-stake network through two hard forks, with the related fixes privately deployed before public disclosure. According to a forum post released on Wednesday, the team packaged the fixes into the Austin hard fork of the Bor client and the Kyoto hard fork of the Heimdall client, both of which followed the standard process for fixing issues that affect consensus: first validated on the Amoy testnet, and then publicly disclosed once the mainnet was activated and the network was secure.The Austin fork fixed two denial-of-service paths in block processing, including a vulnerability where malicious block producers could crash peer nodes by filling them with oversized field data. The Kyoto fork addressed a broader range of consensus hardening issues, with the most severe vulnerability allowing an attacker to force the entire validator set to perform costly and coordinated work with just one crafted transaction—the cost of constructing the transaction is low, but the network processing cost is high. Polygon emphasized that none of the vulnerabilities were observed to be exploited on the mainnet and have been proactively addressed. The two upgrades are now mandatory for node operators and have taken effect without the need for state migration or resynchronization.This disclosure comes at a critical transformation period for Polygon, which has completed the migration of the traditional MATIC token to POL as part of a comprehensive overhaul of its network architecture. The news did not boost the price of POL; according to CoinGecko data, POL traded at approximately $0.09983 on Sunday, down 2.3% in 24 hours, down about 6.8% over the past week, and down about 60.8% over the past year, with a market capitalization of approximately $1.07 billion.

Polygon disclosed multiple previously undisclosed security vulnerabilities recently fixed through two hard forks, Austin and Kyoto, which could affect its proof-of-stake network. The vulnerabilities involve Polygon's Bor and Heimdall clients, including denial-of-service risks, validator resource exhaustion, and defects affecting checkpoint and milestone processing.Polygon stated that the fixes were privately tested before being deployed to the mainnet and publicly disclosed. The most severe issue involved Heimdall, where carefully crafted transactions could force validators to perform excessive processing work, disrupting the network. The Austin hard fork also fixed two denial-of-service risks in Bor that could slow down block processing or cause node crashes. Polygon indicated that these vulnerabilities were not exploited on the mainnet, and the fixes were proactively completed before the details of the vulnerabilities were made public.

The Cronos network associated with Crypto.com has paused operations after detecting an attack on the lending protocol Tectonic. The Cronos Network stated on the X platform that it has identified vulnerabilities on Tectonic and has paused the network. Tectonic also confirmed that it is investigating the related incident and advised users not to interact with the protocol until safety is confirmed. According to DefiLlama data, Tectonic had a total locked value of approximately $121.7 million before the incident, with active loans of about $82.7 million.On-chain researcher Weilin Li attributed the attack to price manipulation of the TONIC token. The attacker bought enough TONIC within 20 minutes to inflate its price by about 100 times, and then used the inflated tokens as collateral to borrow other assets from Tectonic, a method similar to the 2022 Mango Markets oracle manipulation attack. Li initially estimated that the attacker profited about $66 million, and later discovered another address controlled by a different attacker containing about $8 million, bringing the total estimate to around $75 million. Li also stated that the attacker only successfully bridged about $6 million to Ethereum, as the pause on the Cronos network prevented most affected assets from flowing out.Crypto.com CEO Kris Marsalek stated that the company's app and exchange were not attacked, and its security team is assisting Tectonic with the investigation.