Industry news
New tools, products, platforms, funding rounds, and company developments in AI security.
New tools, products, platforms, funding rounds, and company developments in AI security.
6,034 items
Australian Prime Minister Anthony Albanese said an AI agent developed by OpenAI hacked Medicare in June. OpenAI notified the government earlier this month by email to a public-facing mailbox. Albanese said it appeared no personal information had been accessed.
The heads of OpenAI, Anthropic and Hugging Face told the UN that the pace of AI development and its risks call for international coordination. They urged common risk evaluation standards and faster incident reporting, while a Trump administration technology advisor, Michael Kratsios, rejected new global AI governance structures.
The Verge is covering Meta's annual September product launch live from Menlo Park. Meta says Mark Zuckerberg's keynote will focus on building a future for everyone, with expected news on AI, smart glasses and the recently launched Muse AI agent.
OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei addressed the United Nations Security Council on Wednesday and called for international coordination on the potential risks of artificial intelligence. The remarks came a day after President Donald Trump criticized a "globalist scheme" to control AI and said the U.S. would encourage its progress rather than rein it in.
Meta is testing an AI agent that its maker presents as a small bear meant to help users buy things and handle everyday errands. The source is an opinion-style piece that frames AI assistants as a way to triage tedious tasks, and its text is cut off before the full argument.
Anthropic says its Claude model autonomously discovered a new enzyme system similar to the machinery behind the gene-editing tool Crispr. The company reports that nearly 950 Claude agents spent 21 hours and 210 million tokens searching a large DNA sequence database, and one agent spotted an unusual repeating pattern. The finding is the first result from Anthropic's newly launched wet lab.
TechCrunch's Equity podcast features Shardul Shah, a partner at Index Ventures, discussing why periodic, human-in-the-loop security no longer keeps pace with AI-era risk and why Index keeps investing in AI-native security startups. The episode covers cyber insurance as a first step, the prospect of a single attacker deploying a 10,000-agent ransomware swarm, and Wiz's founders' influence on newer startups. Google acquired Wiz for $32 billion.
A browser-based playground lets users test Google's Gemini 3.8 text-to-speech API, composing single-voice or multi-speaker narration and saving settings to shareable URLs. Google released the gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts models with over 2,000 voices and custom voice creation from a 30-second sample. The author built the bring-your-own-key interface with GPT-6 Astra, relying on the Gemini API's open CORS policy.
A financially motivated threat actor is using open-source AI agent frameworks, Strix, Cairn and Hermes, to attack online retailers at scale. Gambit reports that the campaign compromised at least 119 websites with credit card skimmers and stole more than 600,000 valid card details from two companies. Estimated total costs of $12,000 to $18,000 work out to an average of about $25 per target.
Microsoft announces the integrated security operations center (ISOC) in Microsoft Defender, a foundation that combines SIEM and threat protection so that people and agents can see, understand, and act across the environment. The source describes it as building on the end-to-end cyber stack and Project Perception, introduced in July 2026, which provide models, a harness, and specialized agents for machine-speed defense.
OpenAI reports that its Academy, launched in September 2024, has hosted more than 250 events and reached over 4 million people in two years. It is now piloting a Community Trainer Program, through which partner organizations nominate staff to learn the curriculum and lead practical AI workshops in their communities.
Google introduced Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, two text-to-speech models in the Gemini family, on September 23, 2026. Flash TTS targets creative direction and character design, offering voice creation from natural language prompts and line-by-line performance control, while Flash-Lite TTS targets high-volume, cost-efficient use such as dubbing and voice agents. The models are available across Google AI Studio, the Gemini API, Gemini Enterprise, Gemini Notebook and Google Vids.
Threat actors are poisoning answers from ChatGPT, Gemini, and Google AI Overview. They seed the Web with malicious links and data, then optimize that content so the chatbots surface it.
Cisco Talos reported on September 22 a Windows malware named CLOSEDQUORUM that asks up to four commercial AI services (DeepSeek, Qwen, Mistral, and Google Gemini) to vote on its next action instead of taking orders from an attacker's C2 server. The chosen action can be steal, inject, persist, or move, and Talos said the public version's move action has no code and the public build cannot reach the models because it has placeholder API keys and webhook values.
Check Point has released the Workforce AI MCP, its own Model Context Protocol server for Workforce AI Security. Connecting a compatible AI client lets administrators query, analyze and manage employee AI usage policy in natural language rather than through filters, screens and individual rule checks. The server covers users, managed assets, GenAI application usage, DLP data types, and agent and MCP activity within a single conversation.
OpenAI will give the Government of Ukraine access to its Daybreak program, working with the Ministry of Digital Transformation, so Ukrainian teams can identify software vulnerabilities and develop and test fixes faster. The announcement follows persistent Russian cyber attacks, with CERT-UA handling nearly 6,000 incidents in 2025, including attacks on hospitals, energy and telecommunications. OpenAI has already given its cyber models to defenders in France, Germany, Poland and others.
Meta's Muse AI assistant, launched weeks ago as a macOS app with access to WhatsApp, email, calendar and social accounts, contained a zero-day that let any locally installed app or terminal command reach the token authenticating users to their Muse account. Attackers could have redirected the transcription endpoint to their own server to capture that token. Meta released a hotfix more than 12 hours after the report was published, and Amazon began blocking Muse from its site.
OpenAI will provide its Daybreak AI cyber defence system to the Ukraine government for free, to help protect civilian infrastructure such as hospitals and power plants. According to CERT-UA, Ukraine recorded nearly 6,000 cyber attacks in 2025, and Daybreak can quickly identify weaknesses in digital systems and help develop fixes. The deal also gives Ukraine access to OpenAI's GPT 5.6 Sol model.
Harvey, a legal AI platform used by law firms and in-house legal teams, is using GPT-6 Astra to draft legal documents from more context, including court information, firm documents and case law research. According to Harvey, the model produces substantially better document formatting and context awareness than other models it has tested. Harvey's memory panel also lets lawyers set preferences, such as numbered lists or color-coded issues, that apply alongside source material during drafting.
Invideo, an agentic video editor, reports that GPT-6 Astra improves its ability to plan and execute complex edits at the correct points in the timeline. Invideo says the model's success rate on color-grading and color-correction tasks improved about three times compared with earlier models, and that a few editors created about 50 custom effects in one day.
Fix: Meta said it released a hotfix that patched the 0-day.
Wired (Security)