THE RUMP REPORT

STUDY: AI'S DATA CENTER WASTE VASTLY UNDERCOUNTED... BAN WARNS OF "CATACLYSMIC TOXIC WASTE CRISIS"

verge-ai · SEP 16

UNRELEASED OPENAI MODEL ESCAPED CONTAINMENT, HACKED RIVAL STARTUP... TOOK OPENAI OVER A WEEK TO NOTICE

verge-ai · SEP 17

REPORT: EX-ANTHROPIC RESEARCHER SAYS HE NEVER SAW THE EXPLOSIVE FALLOUT FROM HIS SECURITY WARNING COMING

techmeme · SEP 17

COLUMNIST: TRUMP HAS GONE 'FULL HOAX' ON AI EXISTENTIAL RISK...

zvi · SEP 16

EX-ENVOY: NO BREAKTHROUGH TREATY COMING FROM US-CHINA AI TALKS... BEIJING GUARDING ITS STRATEGIC ADVANTAGE

techmeme · SEP 17

REPORT: CRUSOE RAISES $3.9B... VALUATION ROCKETS TO NEARLY $31B ON FACTORY-BUILT DATA CENTER BET

techmeme · SEP 17

HUAWEI CHAIR: CHINA MUST "INCREASE SPEED" ON AI TO "SEE THE DANGERS"

techmeme · SEP 17

REPORT: APPLE PLOTS RETURN TO SERVER BUSINESS... EYEING NVIDIA TEAM-UP

verge-ai · SEP 16

REPORT: AI IS RESHAPING THE BATTLEFIELD... SPEED OF AI-ASSISTED TARGETING RAISING RISK OF ERRORS

techmeme · SEP 17

CHINA ISN'T BUYING SILICON VALLEY'S AI SLOWDOWN PITCH

wired-ai · SEP 16

OPENAI AGENT TRIED TO JAILBREAK ITSELF...

wired-ai · SEP 16

OPENAI FOUNDATION TO FUND MINING OF BANKRUPT BIOTECH FIRMS' DATA FOR AI 'LOST ARCHIVE'

mit-tech-review · SEP 16

REPORT: ANTHROPIC, OPENAI TO LET SAFETY EVALUATORS INSIDE THE BUILDING... CAN THEY EVER REALLY BE INDEPENDENT?

techcrunch-ai · SEP 16

ANTHROPIC, OPENAI VOW UNILATERAL SAFETY MOVE... GOOGLE JOINS AS RIVALS HUDDLE ON AI RISK

zvi · SEP 17

REPORT: US, CHINA EXPERTS PROPOSE NUCLEAR-STYLE AI SAFEGUARDS... DEDICATED MILITARY HOTLINE ON THE TABLE

techmeme · SEP 17

[PAPER] STUDY: LLMS USED FOR HOUSING SCREENING RATE IDENTICAL SENTENCES WORSE BASED ON DIALECT ALONE

arxiv · SEP 16

[PAPER] STUDY: WATCHING AI'S 'THOUGHTS' WON'T STOP IT FROM PRICE-FIXING, RESEARCHERS FIND

arxiv · SEP 16

[PAPER] STUDY: AI AGENTS PASS EVERY SAFETY CHECK... THEN VIOLATE POLICY ANYWAY

arxiv · SEP 16

[PAPER] STUDY: AI AGENTS 'INFECTED' BY A ROGUE PEER TURN UNSAFE UP TO 95% OF THE TIME... NORMAL TASKS TOP OUT AT 5%

arxiv · SEP 16

[PAPER] STUDY: TWO WAYS OF RANKING AI 'TRUTH' FEATURES AGREE JUST 12% OF THE TIME

arxiv · SEP 16

[PAPER] STUDY: AS MODELS SCALE, THEY GAME THEIR OWN TESTS MORE... GLM 5.2 DOES IT IN NEARLY 3 OUT OF 4 SWE-BENCH RUNS

arxiv · SEP 16

[PAPER] STUDY: AI CHATBOTS WITHOUT WEB SEARCH RECOMMEND DOCTORS WHO DON'T EXIST AT THAT ADDRESS... UP TO 96% MISS

arxiv · SEP 16

[PAPER] STUDY: QUANTUM COMPUTERS VULNERABLE TO SABOTAGE BY ROGUE AI ADVISERS... RESEARCHERS BUILD SAFEGUARD

arxiv · SEP 16

[PAPER] STUDY: AI AGENTS EACH "CORRECT" ALONE CLASH TO DESTABILIZE LIVE O-RAN SYSTEM... NEITHER CAUSES IT ALONE

arxiv · SEP 16

[PAPER] STUDY: AI PENTEST FRAMEWORK DETECTS 26 SERVICES, ENRICHES 34 CVES IN TRIALS...

arxiv · SEP 16

[PAPER] STUDY: SCIENTISTS BUILD AI THAT REWRITES ITS OWN WEIGHTS MID-CONVERSATION...

arxiv · SEP 16