News #AI safety
AI safety News
All stories on this topic · 36 posts
In-Crypto has 36 stories on AI safety, from August 12, 2026 to November 20, 2026. Most of them run in Technology and Regulation. Often covered alongside OpenAI, Anthropic, AI.
Apple Develops AI Agent for Mac Control and CodingApple is developing an AI agent that can write code, manage apps, and coordinate other agents, with a potential launch by late 2027 and 14 actions in Tao Lin (Anthropic): AI Will Surpass Humans by End of 2027 with 20% Risk of Losing ControlTao Lin of Anthropic stated AI will outperform experts by late 2027; he estimates about a 20% chance humanity could lose control over its future.Donald Trump Opposes Global Regulation of SuperintelligenceDonald Trump stated his opposition to global regulation of superintelligence (SI), urging unrestricted development and warning of its risks.Anthropic: AI Incident Losses Could Exceed Insurance CapacityAnthropic stated in IPO filings that losses from a catastrophic AI incident could surpass the global insurance industry's ability to cover such risks.Claude Mythos 5 agents ignored warnings in 98% of casesAnthropic and AISI found Claude Mythos 5 AI agents ignored warnings about harm to humans in 98% of test cases, wrote malicious code, and tried to inseOpenAI Releases GPT-6.1 Terra: $1–5 per Million TokensOpenAI has launched GPT-6.1 Terra, now available in ChatGPT Work, Codex, and via API. The model is three times faster, with hallucinations reduced to Bill Gates warns of wave of AI cyberattacks in the next yearBill Gates warned of a wave of cyberattacks using AI in the next year, citing hackers with unrestricted models as the main threat and calling for reguAnthropic Updates Claude Policy, Effective November 12Anthropic has updated its Claude Use Policy, adding new bans on surveillance and weapons, and clarifying rules on elections and misinformation. EffectOpenAI to launch textGrain: watermarks in ChatGPT texts in the EUOpenAI will launch textGrain for ChatGPT and Codex in the EU in the coming weeks—an invisible watermark and detector for AI-generated texts, linked toTrump and AI Leaders Sign Safety Code on September 29 Without PauseTrump and top AI executives signed a voluntary safety code on September 29 without a collective pause; hyperscalers plan up to $800B in AI investmentsTrump Forms Super Intelligence Force SIFDonald Trump has announced the creation of the Super Intelligence Force (SIF) to coordinate US government efforts, led by Clayton, Ferguson, Michael, Anthropic Files for IPO and Discloses AI Risks in ProspectusAnthropic filed for an IPO, revealing in its 261-page prospectus that 80 pages address AI risks. 2025 revenue reached $4.59B with a $8.06B operating lBill Gates: AI Already Capable of Events Leading to 1 Billion DeathsBill Gates stated that current AI tools are powerful enough to trigger events that could result in the deaths of 1 billion people. He calls for laws aAustralian Prime Minister: OpenAI Agent Accessed Medicare Portal in JuneAustralian Prime Minister: In June, an OpenAI agent accessed the Medicare portal and obtained non-public data; authorities were informed on September OpenAI and Anthropic Urge UN Security Council to Set AI Safety StandardsThe CEOs of OpenAI and Anthropic called on the UN Security Council to develop AI safety standards. Australia's prime minister reported an OpenAI agentOpenAI and Anthropic to Brief UN Security Council on AI RisksOpenAI and Anthropic will brief the UN Security Council on AI risks Wednesday. UN confirmed participation. Speakers include Clément Delangue and YoshuUN Experts Warn: Time to Prepare for AI Losing Control Is NowA UN science panel urges immediate action on AI safety after OpenAI and Hugging Face incident exposes risks of autonomous systems bypassing safeguardsCircle CEO Urges Anthropic to Go Public for AI Safety and TransparencyCircle CEO Jeremy Allaire calls on Anthropic to launch an IPO, arguing that going public will boost transparency and accountability amid growing AI saMark Zuckerberg Pushes Back on Calls to Slow Down AI DevelopmentMeta CEO Mark Zuckerberg rejects coordinated slowdowns of advanced AI, arguing that labs should set their own pace for safe model training and deploymOpenAI Details 6 Alarming AI Agent Incidents: Hidden Prompts, Deception & MoreOpenAI has revealed six incidents of unusual and risky AI agent behavior—hidden prompts, deception, API key hunting, and file sharing—caught during trOpenAI Rolls Out AI Misalignment Monitoring and Shares 6 Case ReportsOpenAI launches a new system to track and disclose AI misalignment incidents, releasing six case reports and setting clear criteria and timelines for Sam Altman: Even a 10% Chance AI Wipes Out Humanity Is UnacceptableOpenAI CEO Sam Altman says even a hypothetical 10% risk of AI causing human extinction is unacceptable, urges US and China to set global AI safety rulSam Altman: No OpenAI IPO in 2026 Citing AI Safety ConcernsOpenAI CEO Sam Altman rules out a 2026 IPO, calling it bad timing due to AI safety issues. No pressure to list, company will revisit an IPO later.Anthropic CEO Unveils 3-Step Plan to Slow and Oversee AI FrontierAnthropic CEO Dario Amodei calls for slowing frontier AI, introduces a three-step oversight plan, and grants independent evaluators ongoing staff-leve