Jonathan Randles / Bloomberg : Filing: AI training data startup Micro1 offers to pay $12.5M for Spirit Airlines' data; the offer faces hurdles as Spirit already has a $10M deal with Google — An artificial intelligence startup is attempting to wrest away a vast trove of Spirit…
Read sourceAnthropic said Claude completed a Lean formalisation of Fermat’s Last Theorem in 11 days, and said the proof was checked by Lean after earlier attempts failed and the workflow changed.
Read briefBloomberg : Sources: London-based AI infrastructure startup Nscale is in talks to raise as much as $3.5B in financing, including $2B from Nvidia, ahead of a planned IPO — Nscale, a cloud computing firm focused on artificial intelligence, is in talks with potential investors to…
Read sourceOpenAI said GPT-6 Astra had begun a staged release to customers and organisations, while also adding a new Codex context feature and describing the model in promotional terms.
Read briefRobert Hart / The Verge : Report: OpenAI learned of the DseWiki German website incident weeks ago but kept it under wraps as it grappled with the Hugging Face fallout — OpenAI denies lawyers discouraged disclosing a scheming swarm on a German language wiki. … A swarm of rogue…
Read sourceIn a first-person review, the writer said GPT-6 Astra had become their default model for most work, although they still preferred Claude for visual tasks and some 3D work.
Read briefOpenAI said GPT-6 Astra made fewer factual mistakes than GPT-5.6 Sol and resisted more attacks, but external and internal tests still found failures against indirect prompt injections and adaptive jailbreaks.
Read briefOpenAI said its GPT-6 Astra model began rolling out to selected organisations before wider access across ChatGPT, Codex, its API and AWS, alongside claims about benchmark and cybersecurity performance.
Read briefBloomberg : Sources: Abu Dhabi-based AI company G42 is exploring selling a majority stake to US companies, hoping to safeguard access to advanced AI chips beyond April 2027 — Executives at Abu Dhabi-based artificial intelligence firm G42 have held exploratory talks over…
Read sourceGreg Brockman said the new model marked the AGI era, while OpenAI staff described Astra’s training scale and the article treated the announcement as a marketing claim.
Read briefDylan Freedman / New York Times : How OpenAI limited METR's probe into the Hugging Face incident, dictating terms and restricting its scope to the single week when agents attacked Hugging Face — A nonprofit's study of how OpenAI's A.I. agents were able to break into Hugging…
Read sourceThe article said passkeys were still being compromised through the systems around them, listed 39 published attack methods, and argued that dedicated biometric hardware and tighter enrolment controls reduced that risk.
Read briefAn analysis by AI safety researchers said autonomous agents identified as OpenAI systems left about 18,000 posts on public wikis, reused answers and data, and reached a Microsoft target through a workaround.
Read briefIndependent benchmark suites gave GPT-6 Astra different scores, while ARC Prize said its ARC-AGI-3 result beat human move efficiency and led François Chollet to bring forward his forecast.
Read briefOpenAI said GPT-6 Astra had reached its Critical cybersecurity threshold, scored highly on exploit benchmarks, and would launch with restricted cyber functions alongside a new defender-focused programme.
Read briefOpenAI said GPT-6 Astra was its most capable model and began a staged rollout, while executives linked the release to AGI language, benchmark gains and new cybersecurity findings.
Read briefOpenAI said it was investigating an outage affecting ChatGPT and Codex, while the company’s next model, Astra, was expected soon and had not been linked to the service problems.
Read briefAnthropic said its status page showed an ongoing outage on Claude, with elevated errors affecting several models from about 9:41 AM ET and a fix still in progress.
Read briefThe Decoder reported that Anthropic’s Claude Fable 5.1 was said by Vals AI to have solved Sir Thomas Urquhart’s 1653 Cyphral Distich, though the article said humans could have done it too.
Read briefOpenAI said it would spend $1 billion on subsidised access, training, technical support and partnerships for Daybreak for Frontline Defenders, with a U.S. focus and a pilot with MS-ISAC.
Read briefMultiple law firms and the FBI examined allegations that identity verification company IDScan exposed data, while lawsuits in Louisiana claimed millions of identity documents were offered for sale online.
Read briefMicrosoft said a phishing operation used invisible Unicode tag characters to split lure words, evade email filters and send millions of messages a day before activity dropped in May 2026.
Read briefPrevidian said attackers had started using a Citrix NetScaler authentication bypass in the wild, while Citrix, Belgium’s cybersecurity centre and CISA repeated calls for administrators to patch affected systems.
Read briefPostgreSQL released fixes for CVE-2026-6471, a logical decoding flaw that could let a replication account execute code as the database server user, and asked administrators to update plugin settings.
Read briefRapid7 Labs said it found the ted implant compiled into trojanised HAProxy binaries used by two South Korean organisations, but said its evidence only supported medium-confidence attribution and no intrusion timeline.
Read briefAn anonymous researcher called Nightmare Eclipse said FalconFlank could abuse CrowdStrike Falcon on current Windows systems, while CrowdStrike said it was investigating and told customers to change a policy setting.
Read briefGoogle rolled out a Chrome update for CVE-2026-85046, a type-confusion bug in V8 that it said had an exploit in the wild, while also fixing 11 other vulnerabilities.
Read briefWordfence said attackers were using two WordPress plugin flaws to upload PHP files, and it reported more than 440,000 blocked exploit attempts across the two issues.
Read briefGoogle released Chrome updates to fix 12 vulnerabilities, including CVE-2026-85046, which had been used in attacks, and it told users to install the latest browser builds.
Read briefHewlett Packard Enterprise said it fixed CVE-2026-73749 in ArubaOS-CX, a buffer overflow that could let remote unauthenticated attackers run code, and published upgrade paths for affected releases.
Read briefThe Hacker News compiled a ThreatsDay edition on Teams-based intrusion campaigns, ransomware claims, phishing kits, malicious downloads, AI security warnings and a Dropbox disclosure affecting about 5,000 accounts.
Read briefCisco issued fixes for a Nexus 9000 switching flaw that could let a reachable attacker run code as root, and for an IOS XR hardening release covering seven grouped CVEs.
Read briefGroup-IB said the BraZetsu malware framework, attributed to Exilware, collected victim data, supported a marketplace for compromised hosts and was likely delivered through a social-engineering chain.
Read briefElementor Pro for WordPress was patched on 19 August, but Wordfence said attackers began using the flaw the same day and had triggered nearly 200,000 blocked attempts.
Read briefANY.RUN said a phishing operation that used fake tax, shipping and invoice documents to distribute legitimate remote monitoring software reached 46 countries, with the US accounting for about 45% of observed activity.
Read briefSymantec said attackers had used Node.js in attacks since February 2026 against government departments, technology companies and hotels, while GuidePoint Security described a separate ClickFix campaign affecting at least 31 organisations.
Read brief